Test-set results on an 80/20 split, auto-navigation and full-screen layout

#4
by F4RUKYLDRM - opened
This view is limited to 50 files because it contains too many changes. See the raw diff here.
Files changed (50) hide show
  1. README.md +1 -0
  2. app.py +104 -54
  3. biolmnet/training.py +68 -49
  4. biolmnet/ui/components.py +3 -3
  5. biolmnet/ui/styles.py +21 -0
  6. diagnostics/2026-10-30-other-datasets/results/configs_GBM.json +59 -0
  7. diagnostics/2026-10-30-other-datasets/results/configs_common.json +5 -0
  8. diagnostics/2026-10-30-other-datasets/results/configs_scTrioseq2_top.json +35 -0
  9. diagnostics/2026-10-30-other-datasets/results/grid_COAD_seed42.jsonl +108 -0
  10. diagnostics/2026-10-30-other-datasets/results/grid_COAD_shard0.log +55 -0
  11. diagnostics/2026-10-30-other-datasets/results/grid_COAD_shard1.log +55 -0
  12. diagnostics/2026-10-30-other-datasets/results/grid_COAD_summary.txt +62 -0
  13. diagnostics/2026-10-30-other-datasets/results/grid_GBM_seed42.jsonl +108 -0
  14. diagnostics/2026-10-30-other-datasets/results/grid_GBM_shard0.log +55 -0
  15. diagnostics/2026-10-30-other-datasets/results/grid_GBM_shard1.log +55 -0
  16. diagnostics/2026-10-30-other-datasets/results/grid_GBM_summary.txt +62 -0
  17. diagnostics/2026-10-30-other-datasets/results/grid_scTrioseq2_reduced_shard0.log +5 -0
  18. diagnostics/2026-10-30-other-datasets/results/grid_scTrioseq2_reduced_shard1.log +7 -0
  19. diagnostics/2026-10-30-other-datasets/results/grid_scTrioseq2_reduced_shard2.log +5 -0
  20. diagnostics/2026-10-30-other-datasets/results/grid_scTrioseq2_reduced_shard3.log +7 -0
  21. diagnostics/2026-10-30-other-datasets/results/grid_scTrioseq2_seed42.jsonl +43 -0
  22. diagnostics/2026-10-30-other-datasets/results/grid_scTrioseq2_shard0.log +7 -0
  23. diagnostics/2026-10-30-other-datasets/results/grid_scTrioseq2_shard1.log +14 -0
  24. diagnostics/2026-10-30-other-datasets/results/sanity_baselines.txt +15 -0
  25. diagnostics/2026-10-30-other-datasets/results/seeds_GBM.jsonl +48 -0
  26. diagnostics/2026-10-30-other-datasets/results/seeds_GBM_shard0.log +13 -0
  27. diagnostics/2026-10-30-other-datasets/results/seeds_GBM_shard1.log +13 -0
  28. diagnostics/2026-10-30-other-datasets/results/seeds_GBM_shard2.log +13 -0
  29. diagnostics/2026-10-30-other-datasets/results/seeds_GBM_shard3.log +13 -0
  30. diagnostics/2026-10-30-other-datasets/results/seeds_GBM_summary.txt +8 -0
  31. diagnostics/2026-10-30-other-datasets/results/seeds_scTrioseq2.jsonl +19 -0
  32. diagnostics/2026-10-30-other-datasets/results/seeds_scTrioseq2_common_shard0.log +5 -0
  33. diagnostics/2026-10-30-other-datasets/results/seeds_scTrioseq2_common_shard1.log +4 -0
  34. diagnostics/2026-10-30-other-datasets/results/seeds_scTrioseq2_common_shard2.log +5 -0
  35. diagnostics/2026-10-30-other-datasets/results/seeds_scTrioseq2_common_shard3.log +5 -0
  36. diagnostics/2026-10-30-other-datasets/results/seeds_scTrioseq2_top.jsonl +2 -0
  37. diagnostics/2026-10-30-other-datasets/results/seeds_scTrioseq2_top_shard0.log +1 -0
  38. diagnostics/2026-10-30-other-datasets/results/seeds_scTrioseq2_top_shard1.log +1 -0
  39. diagnostics/2026-10-30-other-datasets/scripts/build_workspace.py +26 -0
  40. diagnostics/2026-10-30-other-datasets/scripts/grid.py +71 -0
  41. diagnostics/2026-10-30-other-datasets/scripts/grid_summary.py +31 -0
  42. diagnostics/2026-10-30-other-datasets/scripts/sanity_baselines.py +32 -0
  43. diagnostics/2026-10-30-other-datasets/scripts/seed_check.py +51 -0
  44. diagnostics/2026-10-30-other-datasets/scripts/seed_summary.py +23 -0
  45. diagnostics/2026-11-06-test-split-reproduction/README.md +36 -0
  46. diagnostics/2026-11-06-test-split-reproduction/results/BRCA.jsonl +30 -0
  47. diagnostics/2026-11-06-test-split-reproduction/results/BRCA_shard0.log +31 -0
  48. diagnostics/2026-11-06-test-split-reproduction/results/BRCA_summary.txt +7 -0
  49. diagnostics/2026-11-06-test-split-reproduction/results/COAD.jsonl +30 -0
  50. diagnostics/2026-11-06-test-split-reproduction/results/COAD_shard0.log +31 -0
README.md CHANGED
@@ -9,6 +9,7 @@ python_version: "3.12"
9
  suggested_hardware: zero-a10g
10
  license: cc-by-nc-4.0
11
  pinned: false
 
12
  ---
13
 
14
  # BioLM-NET Workbench
 
9
  suggested_hardware: zero-a10g
10
  license: cc-by-nc-4.0
11
  pinned: false
12
+ header: mini
13
  ---
14
 
15
  # BioLM-NET Workbench
app.py CHANGED
@@ -34,7 +34,8 @@ from biolmnet.data import (
34
  validate_and_align_omics,
35
  )
36
  from biolmnet.training import (
37
- DEFAULT_VALIDATION_FRACTION,
 
38
  Hyperparameters,
39
  ModelBundle,
40
  pathway_importance,
@@ -173,7 +174,7 @@ def _run_state_rows(active: str, workspace, bundle, run_meta: dict, align: dict,
173
  ("Dataset", dataset, None),
174
  ("Architecture", arch_text, arch_tone),
175
  ("Model", model_text, model_tone),
176
- ("Val accuracy", f"{bundle.metrics['accuracy']:.3f}" if bundle else "—", None),
177
  ("Artifact", "Ready" if bundle else "None", "accent" if bundle else "muted"),
178
  ]
179
  if active == "predict":
@@ -202,8 +203,8 @@ def _run_state_rows(active: str, workspace, bundle, run_meta: dict, align: dict,
202
  f"{bundle.metrics['epochs_completed']}" if bundle else "—",
203
  None,
204
  ),
205
- ("Val accuracy", f"{bundle.metrics['accuracy']:.3f}" if bundle else "—", "accent" if bundle else None),
206
- ("Macro F1", f"{bundle.metrics['f1_macro']:.3f}" if bundle else "—", None),
207
  (
208
  "Predictions",
209
  f"{predicted['n_samples']:,} scored" if predicted.get("n_samples") else "None yet",
@@ -500,7 +501,7 @@ def _loss_plot(history: list[dict[str, float]]) -> go.Figure:
500
  )
501
  figure.add_trace(
502
  go.Scatter(
503
- x=frame["epoch"], y=frame["validation_loss"], mode="lines", name="Validation",
504
  line={"color": "#2c455d", "width": 2.5},
505
  )
506
  )
@@ -517,11 +518,14 @@ def _loss_plot(history: list[dict[str, float]]) -> go.Figure:
517
  return figure
518
 
519
 
520
- def _validation_split_note(validation_fraction: float) -> str:
521
- validation_percent = round(float(validation_fraction) * 100)
 
522
  return (
523
- f'<div class="field-hint">{validation_percent}% of samples are held out for validation '
524
- f"(stratified by class). The remaining {100 - validation_percent}% are used for training.</div>"
 
 
525
  )
526
 
527
 
@@ -534,7 +538,7 @@ def estimate_training_duration(
534
  dropout: float,
535
  projection_dim: int,
536
  fusion_dim: int,
537
- validation_fraction: float,
538
  optimizer: str,
539
  class_weighting: bool,
540
  progress=None,
@@ -587,14 +591,14 @@ def _architecture_audit(bundle: ModelBundle) -> pd.DataFrame:
587
  return pd.DataFrame(rows)
588
 
589
 
590
- def _macro_auc(validation_predictions: pd.DataFrame, label_names: list[str]) -> float | None:
591
- """Macro-averaged one-vs-rest AUC, computed here from the stored validation
592
  probabilities — presentation-layer only; does not change what train()
593
  computes or returns."""
594
  try:
595
  probability_columns = [f"P({label})" for label in label_names]
596
- y_true = validation_predictions["observed"]
597
- y_score = validation_predictions[probability_columns].to_numpy()
598
  y_true_indices = y_true.map({label: index for index, label in enumerate(label_names)}).to_numpy()
599
  if len(label_names) < 2 or len(set(y_true_indices)) < 2:
600
  return None
@@ -615,7 +619,7 @@ def train_workspace(
615
  dropout: float,
616
  projection_dim: int,
617
  fusion_dim: int,
618
- validation_fraction: float,
619
  optimizer: str,
620
  class_weighting: bool,
621
  progress=gr.Progress(track_tqdm=False),
@@ -637,7 +641,7 @@ def train_workspace(
637
  parameters = Hyperparameters(
638
  epochs=int(epochs), batch_size=int(batch_size), learning_rate=float(learning_rate),
639
  weight_decay=float(weight_decay), dropout=float(dropout), projection_dim=int(projection_dim),
640
- fusion_dim=int(fusion_dim), validation_fraction=float(validation_fraction), optimizer=optimizer,
641
  class_weighting=bool(class_weighting),
642
  )
643
 
@@ -650,12 +654,13 @@ def train_workspace(
650
  bundle = result.bundle
651
  artifact_path = save_bundle(bundle)
652
  metrics = bundle.metrics
653
- macro_auc = _macro_auc(result.validation_predictions, bundle.label_names)
654
 
655
  minutes, seconds = divmod(int(elapsed_seconds), 60)
656
  metrics_html = ui.simple_status_html(
657
- f"<strong>Training complete.</strong> Validation accuracy {metrics['accuracy']:.3f} · "
658
- f"elapsed {minutes:02d}:{seconds:02d}. Best validation checkpoint restored; the "
 
659
  "downloadable artifact includes architecture, preprocessing, weights and metrics."
660
  )
661
  loss_figure = _loss_plot(bundle.history)
@@ -673,7 +678,7 @@ def train_workspace(
673
  "train_loss": bundle.history[-1]["training_loss"],
674
  }
675
  return (
676
- bundle, metrics_html, loss_figure, result.validation_predictions, artifact_path,
677
  importance, architecture_audit, run_meta_update, confusion_html,
678
  )
679
  except Exception as exc:
@@ -756,7 +761,7 @@ def _export_panel(bundle: ModelBundle | None, artifact_path: str | None, run_met
756
  f'<div style="font:600 26px/1.1 var(--font-heading);margin-top:6px">Trained bundle ready</div>'
757
  f'<div style="margin-top:8px;font-size:12.5px;color:var(--color-muted)">'
758
  f'Produced by the run that finished at {html.escape(run_meta.get("finished_at", "—"))} with '
759
- f'validation accuracy {bundle.metrics["accuracy"]:.3f} over {len(bundle.label_names)} classes. '
760
  f'Bundle size {ui.human_bytes(total_size)}.</div>',
761
  extra_class="",
762
  )
@@ -944,12 +949,28 @@ def _pathway_attention_table(bundle: ModelBundle) -> pd.DataFrame:
944
  return importance[["branch", "pathway", "genes", "peak_gene_attention", "attention_entropy"]]
945
 
946
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
947
  def refresh_results(bundle: ModelBundle | None, run_meta: dict):
948
  run_meta = run_meta or {}
949
  if bundle is None:
950
  stats = (
951
  '<div class="metric-strip">'
952
- + "".join(ui.stat_plate("—", label) for label in ("Validation accuracy", "Macro F1", "Macro AUC", "Retained pathways"))
953
  + "</div>"
954
  )
955
  return (
@@ -963,14 +984,15 @@ def refresh_results(bundle: ModelBundle | None, run_meta: dict):
963
  )
964
 
965
  macro_auc = run_meta.get("macro_auc")
 
966
  retained = len(bundle.config["gene_pathways"]) + len(bundle.config["dna_pathways"])
967
  stats = (
968
  '<div class="metric-strip">'
969
  + "".join(
970
  [
971
- ui.stat_plate(f"{bundle.metrics['accuracy']:.3f}", "Validation accuracy"),
972
- ui.stat_plate(f"{bundle.metrics['f1_macro']:.3f}", "Macro F1"),
973
- ui.stat_plate(f"{macro_auc:.3f}" if macro_auc is not None else "—", "Macro AUC"),
974
  ui.stat_plate(f"{retained:,}", "Retained pathways"),
975
  ]
976
  )
@@ -1050,7 +1072,7 @@ with gr.Blocks(title="BioLM-NET Workbench", fill_width=True) as demo:
1050
  with gr.Column(min_width=0):
1051
  gr.HTML('<div class="field-label">Repository dataset</div>')
1052
  example_dataset = gr.Dropdown(
1053
- ["BRCA", "COAD", "GBM", "scTrioseq2"], value="BRCA",
1054
  show_label=False, container=False, elem_id="dataset-select",
1055
  )
1056
  with gr.Column(visible=False) as github_group:
@@ -1168,7 +1190,7 @@ with gr.Blocks(title="BioLM-NET Workbench", fill_width=True) as demo:
1168
  gr.HTML(
1169
  ui.title_block_html(
1170
  "Train Model",
1171
- "Fit the prepared architecture with a stratified validation split and balanced "
1172
  "loss options.",
1173
  )
1174
  )
@@ -1201,7 +1223,7 @@ with gr.Blocks(title="BioLM-NET Workbench", fill_width=True) as demo:
1201
  with gr.Column():
1202
  dropout = gr.Slider(0, 0.8, value=0.3, step=0.05, show_label=False, container=False)
1203
 
1204
- gr.HTML('<div class="mono sect sect-spaced">B · Validation &amp; loss</div>')
1205
  with gr.Row(elem_classes=["srow-row", "first"]):
1206
  with gr.Column(min_width=140):
1207
  gr.HTML(ui.esc("Branch projection"))
@@ -1214,15 +1236,13 @@ with gr.Blocks(title="BioLM-NET Workbench", fill_width=True) as demo:
1214
  fusion_dim = gr.Dropdown([8, 12, 16, 32], value=12, show_label=False, container=False)
1215
  with gr.Row(elem_classes=["srow-row"]):
1216
  with gr.Column(min_width=140):
1217
- gr.HTML(ui.esc("Validation fraction"))
1218
  with gr.Column():
1219
- validation_fraction = gr.Slider(
1220
- 0.1, 0.4, value=DEFAULT_VALIDATION_FRACTION, step=0.05, show_label=False, container=False
1221
- )
1222
- validation_split_note = gr.HTML(_validation_split_note(DEFAULT_VALIDATION_FRACTION))
1223
- validation_fraction.change(
1224
- _validation_split_note, inputs=[validation_fraction], outputs=[validation_split_note]
1225
  )
 
 
1226
  with gr.Row(elem_classes=["srow-row"]):
1227
  with gr.Column(min_width=140):
1228
  gr.HTML(ui.esc("Optimizer"))
@@ -1294,6 +1314,11 @@ with gr.Blocks(title="BioLM-NET Workbench", fill_width=True) as demo:
1294
  )
1295
  with gr.Row(elem_classes=["actbar-row"]):
1296
  gr.HTML('<div class="actbar-note">Bundle written to a temp file on export · not persisted between Space restarts</div>')
 
 
 
 
 
1297
 
1298
  # ── Predict ──────────────────────────────────────────────
1299
  with gr.Column(visible=False, elem_id="page-predict", elem_classes=["workspace-page"]) as predict_page:
@@ -1371,7 +1396,7 @@ with gr.Blocks(title="BioLM-NET Workbench", fill_width=True) as demo:
1371
  ui.title_block_html(
1372
  "Results",
1373
  "Everything the run produced, read in one place: architecture audit, "
1374
- "validation outputs, pathway attention and the scored cohort.",
1375
  )
1376
  )
1377
  results_empty_note = gr.HTML(ui.empty_note_html("Build the biological architecture, train a model, or run prediction to populate this page."))
@@ -1384,7 +1409,7 @@ with gr.Blocks(title="BioLM-NET Workbench", fill_width=True) as demo:
1384
  gr.HTML('<div class="mono sect">Training history</div>')
1385
  results_loss_plot = gr.Plot(show_label=False)
1386
  with gr.Column(scale=6, min_width=0):
1387
- gr.HTML('<div class="mono sect">Confusion matrix · validation split</div>')
1388
  results_confusion = gr.HTML("")
1389
  with gr.Row(elem_classes=["results-grid"], equal_height=False):
1390
  with gr.Column(scale=6, min_width=0):
@@ -1393,9 +1418,9 @@ with gr.Blocks(title="BioLM-NET Workbench", fill_width=True) as demo:
1393
  with gr.Column(scale=6, min_width=0):
1394
  gr.HTML('<div class="mono sect">Pathway attention · top retained</div>')
1395
  results_pathway_table = gr.Dataframe(max_height=TABLE_MAX_HEIGHT, interactive=False, wrap=False, value=pd.DataFrame(), headers=[""], column_count=(1, "dynamic"))
1396
- with gr.Column(visible=False) as results_tables_bottom:
1397
- gr.HTML('<div class="mono sect sect-spaced">Validation predictions</div>')
1398
- results_validation_table = gr.Dataframe(max_height=TABLE_MAX_HEIGHT, interactive=False, wrap=False, value=pd.DataFrame(), headers=[""], column_count=(1, "dynamic"))
1399
  gr.HTML('<div class="mono sect sect-spaced">Predictions and class probabilities</div>')
1400
  results_prediction_table = gr.Dataframe(max_height=TABLE_MAX_HEIGHT, interactive=False, wrap=False, value=pd.DataFrame(), headers=[""], column_count=(1, "dynamic"))
1401
  with gr.Row(elem_classes=["actbar-row"]):
@@ -1433,7 +1458,7 @@ with gr.Blocks(title="BioLM-NET Workbench", fill_width=True) as demo:
1433
  intro_next_buttons.append(next_btn)
1434
  else:
1435
  back_btn = gr.Button("Back", size="sm", variant="secondary", elem_classes=["btn-ghost"])
1436
- start_btn = gr.Button("Start with the BRCA example", size="sm", variant="primary")
1437
  intro_back_buttons.append(back_btn)
1438
  intro_start_buttons.append(start_btn)
1439
  intro_cards.append(intro_card)
@@ -1452,10 +1477,21 @@ with gr.Blocks(title="BioLM-NET Workbench", fill_width=True) as demo:
1452
 
1453
  return _handler
1454
 
 
 
1455
  for button, key in zip(rail_row_buttons, STAGE_KEYS):
1456
- button.click(
1457
- _nav(key), inputs=chrome_inputs_tail, outputs=[active_page_state, *page_columns, *chrome_outputs]
1458
- )
 
 
 
 
 
 
 
 
 
1459
 
1460
  def _intro_updates(index: int, visible: bool = True):
1461
  index = max(0, min(len(ui.INTRO_CARDS) - 1, int(index)))
@@ -1468,13 +1504,13 @@ with gr.Blocks(title="BioLM-NET Workbench", fill_width=True) as demo:
1468
  def _intro_close(dont_show):
1469
  return 0, gr.update(visible=False), bool(dont_show)
1470
 
1471
- def _intro_start_brca(dont_show):
1472
  return (
1473
  0,
1474
  gr.update(visible=False),
1475
  bool(dont_show),
1476
  "BioLM-NET examples",
1477
- "BRCA",
1478
  *_source_visibility("BioLM-NET examples"),
1479
  )
1480
 
@@ -1505,7 +1541,7 @@ with gr.Blocks(title="BioLM-NET Workbench", fill_width=True) as demo:
1505
  )
1506
  for button in intro_start_buttons:
1507
  button.click(
1508
- _intro_start_brca, inputs=[intro_dont_show],
1509
  outputs=[
1510
  intro_step_state, intro_group, intro_dismissed_state,
1511
  source_mode, example_dataset, example_group, github_group, upload_group,
@@ -1570,6 +1606,9 @@ with gr.Blocks(title="BioLM-NET Workbench", fill_width=True) as demo:
1570
  _after_prepare, inputs=[workspace_state, run_meta_state, session_id_state],
1571
  outputs=[prepare_button, revalidate_button, *chrome_outputs],
1572
  show_progress="hidden",
 
 
 
1573
  )
1574
 
1575
  revalidate_button.click(
@@ -1586,6 +1625,9 @@ with gr.Blocks(title="BioLM-NET Workbench", fill_width=True) as demo:
1586
  _after_prepare, inputs=[workspace_state, run_meta_state, session_id_state],
1587
  outputs=[prepare_button, revalidate_button, *chrome_outputs],
1588
  show_progress="hidden",
 
 
 
1589
  )
1590
 
1591
  show_symbols_btn.click(_reveal_symbols, inputs=[run_meta_state], outputs=[unmatched_reveal])
@@ -1631,14 +1673,14 @@ with gr.Blocks(title="BioLM-NET Workbench", fill_width=True) as demo:
1631
  # passed into/out of `train_workspace` itself.
1632
  train_inputs = [
1633
  workspace_state, epochs, batch_size, learning_rate, weight_decay, dropout, projection_dim,
1634
- fusion_dim, validation_fraction, optimizer, class_weighting,
1635
  ]
1636
 
1637
  def _after_train(bundle, run_meta, session_id):
1638
  chrome = _refresh_chrome("train", bundle is not None and True, bundle, run_meta, {}, {}, session_id)
1639
  return _idle("Train BioLM-NET on ZeroGPU", ["actbar-primary-btn"]), *chrome
1640
 
1641
- validation_predictions_state = gr.State(pd.DataFrame())
1642
  importance_state = gr.State(pd.DataFrame())
1643
  architecture_audit_state = gr.State(pd.DataFrame())
1644
  confusion_html_state = gr.State("")
@@ -1650,7 +1692,7 @@ with gr.Blocks(title="BioLM-NET Workbench", fill_width=True) as demo:
1650
  ).then(
1651
  train_workspace, inputs=train_inputs,
1652
  outputs=[
1653
- model_state, training_status, loss_plot, validation_predictions_state, artifact_path_state,
1654
  importance_state, architecture_audit_state, run_meta_update_state, confusion_html_state,
1655
  ],
1656
  show_progress="hidden",
@@ -1677,13 +1719,13 @@ with gr.Blocks(title="BioLM-NET Workbench", fill_width=True) as demo:
1677
  outputs=[results_stats, results_stats_row, results_empty_note, results_content, results_loss_plot],
1678
  show_progress="hidden",
1679
  ).then(
1680
- lambda df: gr.update(value=df), inputs=[validation_predictions_state], outputs=[results_validation_table],
1681
  show_progress="hidden",
1682
  ).then(
1683
- lambda df: gr.update(value=df), inputs=[architecture_audit_state], outputs=[results_architecture_table],
1684
  show_progress="hidden",
1685
  ).then(
1686
- lambda bundle: gr.update(value=_pathway_attention_table(bundle)) if bundle else gr.update(value=pd.DataFrame()),
1687
  inputs=[model_state], outputs=[results_pathway_table],
1688
  show_progress="hidden",
1689
  ).then(
@@ -1692,6 +1734,9 @@ with gr.Blocks(title="BioLM-NET Workbench", fill_width=True) as demo:
1692
  ).then(
1693
  lambda bundle: gr.update(visible=bundle is not None), inputs=[model_state], outputs=[results_tables_bottom],
1694
  show_progress="hidden",
 
 
 
1695
  )
1696
  # training_status is NOT re-set here: train_workspace's own return
1697
  # already carries the final "Training complete · accuracy · elapsed"
@@ -1746,14 +1791,19 @@ with gr.Blocks(title="BioLM-NET Workbench", fill_width=True) as demo:
1746
  outputs=chrome_outputs,
1747
  show_progress="hidden",
1748
  ).then(
1749
- lambda df: gr.update(value=df), inputs=[prediction_table], outputs=[results_prediction_table],
1750
  show_progress="hidden",
1751
  ).then(
1752
  refresh_results, inputs=[model_state, run_meta_state],
1753
  outputs=[results_stats, results_stats_row, results_empty_note, results_content, results_loss_plot],
1754
  show_progress="hidden",
 
 
 
1755
  )
1756
 
 
 
1757
 
1758
  if __name__ == "__main__":
1759
  demo.queue(default_concurrency_limit=1).launch(theme=THEME, css=CSS)
 
34
  validate_and_align_omics,
35
  )
36
  from biolmnet.training import (
37
+ DEFAULT_TEST_FRACTION,
38
+ EARLY_STOPPING_FRACTION,
39
  Hyperparameters,
40
  ModelBundle,
41
  pathway_importance,
 
174
  ("Dataset", dataset, None),
175
  ("Architecture", arch_text, arch_tone),
176
  ("Model", model_text, model_tone),
177
+ (f"{_score_split(bundle)} accuracy" if bundle else "Test accuracy", f"{bundle.metrics['accuracy']:.3f}" if bundle else "—", None),
178
  ("Artifact", "Ready" if bundle else "None", "accent" if bundle else "muted"),
179
  ]
180
  if active == "predict":
 
203
  f"{bundle.metrics['epochs_completed']}" if bundle else "—",
204
  None,
205
  ),
206
+ (f"{_score_split(bundle)} accuracy" if bundle else "Test accuracy", f"{bundle.metrics['accuracy']:.3f}" if bundle else "—", "accent" if bundle else None),
207
+ (f"{_score_split(bundle)} macro F1" if bundle else "Test macro F1", f"{bundle.metrics['f1_macro']:.3f}" if bundle else "—", None),
208
  (
209
  "Predictions",
210
  f"{predicted['n_samples']:,} scored" if predicted.get("n_samples") else "None yet",
 
501
  )
502
  figure.add_trace(
503
  go.Scatter(
504
+ x=frame["epoch"], y=frame["validation_loss"], mode="lines", name="Validation (early stopping)",
505
  line={"color": "#2c455d", "width": 2.5},
506
  )
507
  )
 
518
  return figure
519
 
520
 
521
+ def _split_note(test_fraction: float) -> str:
522
+ test_percent = round(float(test_fraction) * 100)
523
+ validation_percent = round((1 - float(test_fraction)) * EARLY_STOPPING_FRACTION * 100)
524
  return (
525
+ f'<div class="field-hint">{test_percent}% of samples are held out as an untouched test set '
526
+ f"(stratified by class); every reported score comes from it. Of the remaining "
527
+ f"{100 - test_percent}%, {validation_percent}% of all samples choose the early-stopping checkpoint "
528
+ f"and {100 - test_percent - validation_percent}% train the model.</div>"
529
  )
530
 
531
 
 
538
  dropout: float,
539
  projection_dim: int,
540
  fusion_dim: int,
541
+ test_fraction: float,
542
  optimizer: str,
543
  class_weighting: bool,
544
  progress=None,
 
591
  return pd.DataFrame(rows)
592
 
593
 
594
+ def _macro_auc(test_predictions: pd.DataFrame, label_names: list[str]) -> float | None:
595
+ """Macro-averaged one-vs-rest AUC, computed here from the stored test-set
596
  probabilities — presentation-layer only; does not change what train()
597
  computes or returns."""
598
  try:
599
  probability_columns = [f"P({label})" for label in label_names]
600
+ y_true = test_predictions["observed"]
601
+ y_score = test_predictions[probability_columns].to_numpy()
602
  y_true_indices = y_true.map({label: index for index, label in enumerate(label_names)}).to_numpy()
603
  if len(label_names) < 2 or len(set(y_true_indices)) < 2:
604
  return None
 
619
  dropout: float,
620
  projection_dim: int,
621
  fusion_dim: int,
622
+ test_fraction: float,
623
  optimizer: str,
624
  class_weighting: bool,
625
  progress=gr.Progress(track_tqdm=False),
 
641
  parameters = Hyperparameters(
642
  epochs=int(epochs), batch_size=int(batch_size), learning_rate=float(learning_rate),
643
  weight_decay=float(weight_decay), dropout=float(dropout), projection_dim=int(projection_dim),
644
+ fusion_dim=int(fusion_dim), test_fraction=float(test_fraction), optimizer=optimizer,
645
  class_weighting=bool(class_weighting),
646
  )
647
 
 
654
  bundle = result.bundle
655
  artifact_path = save_bundle(bundle)
656
  metrics = bundle.metrics
657
+ macro_auc = _macro_auc(result.test_predictions, bundle.label_names)
658
 
659
  minutes, seconds = divmod(int(elapsed_seconds), 60)
660
  metrics_html = ui.simple_status_html(
661
+ f"<strong>Training complete.</strong> Test accuracy {metrics['accuracy']:.3f} · "
662
+ f"test macro F1 {metrics['f1_macro']:.3f} · elapsed {minutes:02d}:{seconds:02d}. Scores come from the "
663
+ "untouched test set; the early-stopping checkpoint was chosen on a separate validation slice. The "
664
  "downloadable artifact includes architecture, preprocessing, weights and metrics."
665
  )
666
  loss_figure = _loss_plot(bundle.history)
 
678
  "train_loss": bundle.history[-1]["training_loss"],
679
  }
680
  return (
681
+ bundle, metrics_html, loss_figure, result.test_predictions, artifact_path,
682
  importance, architecture_audit, run_meta_update, confusion_html,
683
  )
684
  except Exception as exc:
 
761
  f'<div style="font:600 26px/1.1 var(--font-heading);margin-top:6px">Trained bundle ready</div>'
762
  f'<div style="margin-top:8px;font-size:12.5px;color:var(--color-muted)">'
763
  f'Produced by the run that finished at {html.escape(run_meta.get("finished_at", "—"))} with '
764
+ f'{_score_split(bundle).lower()} accuracy {bundle.metrics["accuracy"]:.3f} over {len(bundle.label_names)} classes. '
765
  f'Bundle size {ui.human_bytes(total_size)}.</div>',
766
  extra_class="",
767
  )
 
949
  return importance[["branch", "pathway", "genes", "peak_gene_attention", "attention_entropy"]]
950
 
951
 
952
+ def _for_display(frame: pd.DataFrame | None) -> pd.DataFrame:
953
+ """Round float columns to 3 decimals for on-screen tables only; the
954
+ downloadable prediction file keeps full precision."""
955
+ if frame is None or frame.empty:
956
+ return frame if frame is not None else pd.DataFrame()
957
+ float_columns = frame.select_dtypes("float").columns
958
+ # Model outputs are float32; widen before rounding so 0.02 shows as 0.02, not 0.0199999995.
959
+ return frame.astype({column: "float64" for column in float_columns}).round({column: 3 for column in float_columns})
960
+
961
+
962
+ def _score_split(bundle: ModelBundle) -> str:
963
+ """Which partition a bundle's metrics come from. Artifacts saved before the
964
+ 80/20 test split carry validation scores and must stay labelled as such."""
965
+ return "Test" if bundle.metrics.get("evaluation_split") == "test" else "Validation"
966
+
967
+
968
  def refresh_results(bundle: ModelBundle | None, run_meta: dict):
969
  run_meta = run_meta or {}
970
  if bundle is None:
971
  stats = (
972
  '<div class="metric-strip">'
973
+ + "".join(ui.stat_plate("—", label) for label in ("Test accuracy", "Test macro F1", "Test macro AUC", "Retained pathways"))
974
  + "</div>"
975
  )
976
  return (
 
984
  )
985
 
986
  macro_auc = run_meta.get("macro_auc")
987
+ split = _score_split(bundle)
988
  retained = len(bundle.config["gene_pathways"]) + len(bundle.config["dna_pathways"])
989
  stats = (
990
  '<div class="metric-strip">'
991
  + "".join(
992
  [
993
+ ui.stat_plate(f"{bundle.metrics['accuracy']:.3f}", f"{split} accuracy"),
994
+ ui.stat_plate(f"{bundle.metrics['f1_macro']:.3f}", f"{split} macro F1"),
995
+ ui.stat_plate(f"{macro_auc:.3f}" if macro_auc is not None else "—", f"{split} macro AUC"),
996
  ui.stat_plate(f"{retained:,}", "Retained pathways"),
997
  ]
998
  )
 
1072
  with gr.Column(min_width=0):
1073
  gr.HTML('<div class="field-label">Repository dataset</div>')
1074
  example_dataset = gr.Dropdown(
1075
+ ["scTrioseq2", "BRCA", "COAD", "GBM"], value="scTrioseq2",
1076
  show_label=False, container=False, elem_id="dataset-select",
1077
  )
1078
  with gr.Column(visible=False) as github_group:
 
1190
  gr.HTML(
1191
  ui.title_block_html(
1192
  "Train Model",
1193
+ "Fit the prepared architecture on a stratified 80/20 train/test split with balanced "
1194
  "loss options.",
1195
  )
1196
  )
 
1223
  with gr.Column():
1224
  dropout = gr.Slider(0, 0.8, value=0.3, step=0.05, show_label=False, container=False)
1225
 
1226
+ gr.HTML('<div class="mono sect sect-spaced">B · Split &amp; loss</div>')
1227
  with gr.Row(elem_classes=["srow-row", "first"]):
1228
  with gr.Column(min_width=140):
1229
  gr.HTML(ui.esc("Branch projection"))
 
1236
  fusion_dim = gr.Dropdown([8, 12, 16, 32], value=12, show_label=False, container=False)
1237
  with gr.Row(elem_classes=["srow-row"]):
1238
  with gr.Column(min_width=140):
1239
+ gr.HTML(ui.esc("Test fraction"))
1240
  with gr.Column():
1241
+ test_fraction = gr.Slider(
1242
+ 0.1, 0.4, value=DEFAULT_TEST_FRACTION, step=0.05, show_label=False, container=False
 
 
 
 
1243
  )
1244
+ split_note = gr.HTML(_split_note(DEFAULT_TEST_FRACTION))
1245
+ test_fraction.change(_split_note, inputs=[test_fraction], outputs=[split_note])
1246
  with gr.Row(elem_classes=["srow-row"]):
1247
  with gr.Column(min_width=140):
1248
  gr.HTML(ui.esc("Optimizer"))
 
1314
  )
1315
  with gr.Row(elem_classes=["actbar-row"]):
1316
  gr.HTML('<div class="actbar-note">Bundle written to a temp file on export · not persisted between Space restarts</div>')
1317
+ with gr.Row(elem_classes=["actbar-buttons"]):
1318
+ continue_predict_button = gr.Button(
1319
+ "Continue to Predict", size="sm", variant="primary",
1320
+ elem_id="continue-predict-btn", elem_classes=["actbar-primary-btn"],
1321
+ )
1322
 
1323
  # ── Predict ──────────────────────────────────────────────
1324
  with gr.Column(visible=False, elem_id="page-predict", elem_classes=["workspace-page"]) as predict_page:
 
1396
  ui.title_block_html(
1397
  "Results",
1398
  "Everything the run produced, read in one place: architecture audit, "
1399
+ "test-set outputs, pathway attention and the scored cohort.",
1400
  )
1401
  )
1402
  results_empty_note = gr.HTML(ui.empty_note_html("Build the biological architecture, train a model, or run prediction to populate this page."))
 
1409
  gr.HTML('<div class="mono sect">Training history</div>')
1410
  results_loss_plot = gr.Plot(show_label=False)
1411
  with gr.Column(scale=6, min_width=0):
1412
+ gr.HTML('<div class="mono sect">Confusion matrix · test set</div>')
1413
  results_confusion = gr.HTML("")
1414
  with gr.Row(elem_classes=["results-grid"], equal_height=False):
1415
  with gr.Column(scale=6, min_width=0):
 
1418
  with gr.Column(scale=6, min_width=0):
1419
  gr.HTML('<div class="mono sect">Pathway attention · top retained</div>')
1420
  results_pathway_table = gr.Dataframe(max_height=TABLE_MAX_HEIGHT, interactive=False, wrap=False, value=pd.DataFrame(), headers=[""], column_count=(1, "dynamic"))
1421
+ with gr.Column(visible=False, elem_classes=["results-bottom"]) as results_tables_bottom:
1422
+ gr.HTML('<div class="mono sect sect-spaced">Test-set predictions</div>')
1423
+ results_test_table = gr.Dataframe(max_height=TABLE_MAX_HEIGHT, interactive=False, wrap=False, value=pd.DataFrame(), headers=[""], column_count=(1, "dynamic"))
1424
  gr.HTML('<div class="mono sect sect-spaced">Predictions and class probabilities</div>')
1425
  results_prediction_table = gr.Dataframe(max_height=TABLE_MAX_HEIGHT, interactive=False, wrap=False, value=pd.DataFrame(), headers=[""], column_count=(1, "dynamic"))
1426
  with gr.Row(elem_classes=["actbar-row"]):
 
1458
  intro_next_buttons.append(next_btn)
1459
  else:
1460
  back_btn = gr.Button("Back", size="sm", variant="secondary", elem_classes=["btn-ghost"])
1461
+ start_btn = gr.Button("Start with the scTrioseq2 example", size="sm", variant="primary")
1462
  intro_back_buttons.append(back_btn)
1463
  intro_start_buttons.append(start_btn)
1464
  intro_cards.append(intro_card)
 
1477
 
1478
  return _handler
1479
 
1480
+ nav_outputs = [active_page_state, *page_columns, *chrome_outputs]
1481
+
1482
  for button, key in zip(rail_row_buttons, STAGE_KEYS):
1483
+ button.click(_nav(key), inputs=chrome_inputs_tail, outputs=nav_outputs)
1484
+
1485
+ def _advance_to(page_key):
1486
+ """Move to the next workflow page once a step has succeeded; on failure
1487
+ stay put so the error message stays in view."""
1488
+ def _handler(result, workspace, bundle, run_meta, align, predicted, session_id):
1489
+ succeeded = not result.empty if isinstance(result, pd.DataFrame) else result is not None
1490
+ if not succeeded:
1491
+ return tuple(gr.skip() for _ in nav_outputs)
1492
+ return _enter_page(page_key, workspace, bundle, run_meta, align, predicted, session_id)
1493
+
1494
+ return _handler
1495
 
1496
  def _intro_updates(index: int, visible: bool = True):
1497
  index = max(0, min(len(ui.INTRO_CARDS) - 1, int(index)))
 
1504
  def _intro_close(dont_show):
1505
  return 0, gr.update(visible=False), bool(dont_show)
1506
 
1507
+ def _intro_start_example(dont_show):
1508
  return (
1509
  0,
1510
  gr.update(visible=False),
1511
  bool(dont_show),
1512
  "BioLM-NET examples",
1513
+ "scTrioseq2",
1514
  *_source_visibility("BioLM-NET examples"),
1515
  )
1516
 
 
1541
  )
1542
  for button in intro_start_buttons:
1543
  button.click(
1544
+ _intro_start_example, inputs=[intro_dont_show],
1545
  outputs=[
1546
  intro_step_state, intro_group, intro_dismissed_state,
1547
  source_mode, example_dataset, example_group, github_group, upload_group,
 
1606
  _after_prepare, inputs=[workspace_state, run_meta_state, session_id_state],
1607
  outputs=[prepare_button, revalidate_button, *chrome_outputs],
1608
  show_progress="hidden",
1609
+ ).then(
1610
+ _advance_to("train"), inputs=[workspace_state, *chrome_inputs_tail], outputs=nav_outputs,
1611
+ show_progress="hidden",
1612
  )
1613
 
1614
  revalidate_button.click(
 
1625
  _after_prepare, inputs=[workspace_state, run_meta_state, session_id_state],
1626
  outputs=[prepare_button, revalidate_button, *chrome_outputs],
1627
  show_progress="hidden",
1628
+ ).then(
1629
+ _advance_to("train"), inputs=[workspace_state, *chrome_inputs_tail], outputs=nav_outputs,
1630
+ show_progress="hidden",
1631
  )
1632
 
1633
  show_symbols_btn.click(_reveal_symbols, inputs=[run_meta_state], outputs=[unmatched_reveal])
 
1673
  # passed into/out of `train_workspace` itself.
1674
  train_inputs = [
1675
  workspace_state, epochs, batch_size, learning_rate, weight_decay, dropout, projection_dim,
1676
+ fusion_dim, test_fraction, optimizer, class_weighting,
1677
  ]
1678
 
1679
  def _after_train(bundle, run_meta, session_id):
1680
  chrome = _refresh_chrome("train", bundle is not None and True, bundle, run_meta, {}, {}, session_id)
1681
  return _idle("Train BioLM-NET on ZeroGPU", ["actbar-primary-btn"]), *chrome
1682
 
1683
+ test_predictions_state = gr.State(pd.DataFrame())
1684
  importance_state = gr.State(pd.DataFrame())
1685
  architecture_audit_state = gr.State(pd.DataFrame())
1686
  confusion_html_state = gr.State("")
 
1692
  ).then(
1693
  train_workspace, inputs=train_inputs,
1694
  outputs=[
1695
+ model_state, training_status, loss_plot, test_predictions_state, artifact_path_state,
1696
  importance_state, architecture_audit_state, run_meta_update_state, confusion_html_state,
1697
  ],
1698
  show_progress="hidden",
 
1719
  outputs=[results_stats, results_stats_row, results_empty_note, results_content, results_loss_plot],
1720
  show_progress="hidden",
1721
  ).then(
1722
+ lambda df: gr.update(value=_for_display(df)), inputs=[test_predictions_state], outputs=[results_test_table],
1723
  show_progress="hidden",
1724
  ).then(
1725
+ lambda df: gr.update(value=_for_display(df)), inputs=[architecture_audit_state], outputs=[results_architecture_table],
1726
  show_progress="hidden",
1727
  ).then(
1728
+ lambda bundle: gr.update(value=_for_display(_pathway_attention_table(bundle))) if bundle else gr.update(value=pd.DataFrame()),
1729
  inputs=[model_state], outputs=[results_pathway_table],
1730
  show_progress="hidden",
1731
  ).then(
 
1734
  ).then(
1735
  lambda bundle: gr.update(visible=bundle is not None), inputs=[model_state], outputs=[results_tables_bottom],
1736
  show_progress="hidden",
1737
+ ).then(
1738
+ _advance_to("export"), inputs=[model_state, *chrome_inputs_tail], outputs=nav_outputs,
1739
+ show_progress="hidden",
1740
  )
1741
  # training_status is NOT re-set here: train_workspace's own return
1742
  # already carries the final "Training complete · accuracy · elapsed"
 
1791
  outputs=chrome_outputs,
1792
  show_progress="hidden",
1793
  ).then(
1794
+ lambda df: gr.update(value=_for_display(df)), inputs=[prediction_table], outputs=[results_prediction_table],
1795
  show_progress="hidden",
1796
  ).then(
1797
  refresh_results, inputs=[model_state, run_meta_state],
1798
  outputs=[results_stats, results_stats_row, results_empty_note, results_content, results_loss_plot],
1799
  show_progress="hidden",
1800
+ ).then(
1801
+ _advance_to("results"), inputs=[prediction_table, *chrome_inputs_tail], outputs=nav_outputs,
1802
+ show_progress="hidden",
1803
  )
1804
 
1805
+ continue_predict_button.click(_nav("predict"), inputs=chrome_inputs_tail, outputs=nav_outputs)
1806
+
1807
 
1808
  if __name__ == "__main__":
1809
  demo.queue(default_concurrency_limit=1).launch(theme=THEME, css=CSS)
biolmnet/training.py CHANGED
@@ -26,9 +26,12 @@ from .model import BioLMNet
26
 
27
  ProgressCallback = Callable[[float, str], None]
28
 
29
- # 75% training / 25% validation. The validation partition drives early
30
- # stopping and best-checkpoint selection, so it is not an untouched test set.
31
- DEFAULT_VALIDATION_FRACTION = 0.25
 
 
 
32
 
33
  # Activations BioLMNet accepts.
34
  ACTIVATIONS = ("relu", "sigmoid", "tanh")
@@ -47,7 +50,8 @@ class Hyperparameters:
47
  dropout: float = 0.3
48
  projection_dim: int = 64
49
  fusion_dim: int = 12
50
- validation_fraction: float = DEFAULT_VALIDATION_FRACTION
 
51
  optimizer: str = "Adam"
52
  class_weighting: bool = True
53
  early_stopping_patience: int = 12
@@ -66,8 +70,10 @@ class Hyperparameters:
66
  raise ValueError("Learning rate must be in (0, 1].")
67
  if not 0 <= self.dropout < 1:
68
  raise ValueError("Dropout must be in [0, 1).")
69
- if not 0.05 <= self.validation_fraction <= 0.5:
70
- raise ValueError("Validation fraction must be between 0.05 and 0.5.")
 
 
71
  if self.optimizer.lower() not in {"adam", "sgd"}:
72
  raise ValueError("Optimizer must be Adam or SGD.")
73
  for name in ("biological_activation", "projection_activation", "fusion_activation"):
@@ -95,7 +101,7 @@ class ModelBundle:
95
  @dataclass
96
  class TrainingResult:
97
  bundle: ModelBundle
98
- validation_predictions: pd.DataFrame
99
  confusion: np.ndarray
100
 
101
 
@@ -107,34 +113,49 @@ def set_reproducible_seed(seed: int) -> None:
107
  torch.cuda.manual_seed_all(seed)
108
 
109
 
110
- def split_train_validation(
111
- labels: np.ndarray, validation_fraction: float, seed: int
112
  ) -> tuple[np.ndarray, np.ndarray]:
113
- """Stratified sample-index split shared by both omics and the labels."""
114
  classes, class_counts = np.unique(labels, return_counts=True)
115
  if class_counts.min() < 2:
116
  raise ValueError(
117
- "Each class needs at least two samples for a stratified train/validation split."
118
  )
119
  sample_count = len(labels)
120
- validation_count = math.ceil(validation_fraction * sample_count)
121
- train_count = sample_count - validation_count
122
- if min(validation_count, train_count) < len(classes):
123
  raise ValueError(
124
- f"A validation fraction of {validation_fraction:.2f} on {sample_count} samples "
125
- f"gives {train_count} training and {validation_count} validation samples; "
126
  f"each partition needs at least one sample per class ({len(classes)} classes) "
127
- "for a stratified split. Adjust the validation fraction or add samples."
128
  )
129
 
130
  indices = np.arange(sample_count)
131
- train_index, validation_index = train_test_split(
132
  indices,
133
- test_size=validation_fraction,
134
  random_state=seed,
135
  stratify=labels,
136
  )
137
- return train_index, validation_index
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
138
 
139
 
140
  def _fit_scaler(values: np.ndarray) -> tuple[np.ndarray, np.ndarray]:
@@ -207,8 +228,11 @@ def train(
207
  hyperparameters.validate()
208
  set_reproducible_seed(hyperparameters.seed)
209
  labels = workspace.labels
210
- train_index, validation_index = split_train_validation(
211
- labels, hyperparameters.validation_fraction, hyperparameters.seed
 
 
 
212
  )
213
  gene_mean, gene_scale = _fit_scaler(
214
  workspace.gene_expression[train_index]
@@ -226,8 +250,11 @@ def train(
226
  dna_validation = _scale(
227
  workspace.dna_methylation[validation_index], dna_mean, dna_scale
228
  )
 
 
229
  y_train = labels[train_index]
230
  y_validation = labels[validation_index]
 
231
 
232
  dataset = TensorDataset(
233
  torch.from_numpy(gene_train),
@@ -331,55 +358,47 @@ def train(
331
 
332
  if best_state is not None:
333
  model.load_state_dict(best_state)
334
- validation_loss, probabilities, predictions = _evaluate(
 
335
  model,
336
- gene_validation,
337
- dna_validation,
338
- y_validation,
339
  device,
340
  )
341
  confusion = confusion_matrix(
342
- y_validation,
343
  predictions,
344
  labels=np.arange(len(workspace.label_names)),
345
  )
346
  metrics = {
347
- "validation_loss": validation_loss,
348
- "accuracy": float(accuracy_score(y_validation, predictions)),
 
 
349
  "f1_macro": float(
350
- f1_score(
351
- y_validation, predictions, average="macro", zero_division=0
352
- )
353
  ),
354
  "f1_weighted": float(
355
- f1_score(
356
- y_validation, predictions, average="weighted", zero_division=0
357
- )
358
  ),
359
  "precision_macro": float(
360
- precision_score(
361
- y_validation, predictions, average="macro", zero_division=0
362
- )
363
  ),
364
  "recall_macro": float(
365
- recall_score(
366
- y_validation, predictions, average="macro", zero_division=0
367
- )
368
  ),
369
  "epochs_completed": len(history),
370
  "training_samples": int(len(train_index)),
371
  "validation_samples": int(len(validation_index)),
 
372
  "device": str(device),
373
  }
374
  predictions_frame = pd.DataFrame(
375
  {
376
- "sample_row": validation_index,
377
- "observed": [
378
- workspace.label_names[value] for value in y_validation
379
- ],
380
- "predicted": [
381
- workspace.label_names[value] for value in predictions
382
- ],
383
  }
384
  )
385
  for index, label in enumerate(workspace.label_names):
@@ -422,7 +441,7 @@ def train(
422
  )
423
  return TrainingResult(
424
  bundle=bundle,
425
- validation_predictions=predictions_frame,
426
  confusion=confusion,
427
  )
428
 
 
26
 
27
  ProgressCallback = Callable[[float, str], None]
28
 
29
+ # 80% training / 20% test, as in the BioLM-NET paper. The test partition is never
30
+ # used during fitting; every reported score comes from it.
31
+ DEFAULT_TEST_FRACTION = 0.2
32
+ # Share of the training partition held out to choose the early-stopping checkpoint
33
+ # (1/8 of 80% = 10% of all samples), so checkpoint selection never sees the test set.
34
+ EARLY_STOPPING_FRACTION = 0.125
35
 
36
  # Activations BioLMNet accepts.
37
  ACTIVATIONS = ("relu", "sigmoid", "tanh")
 
50
  dropout: float = 0.3
51
  projection_dim: int = 64
52
  fusion_dim: int = 12
53
+ test_fraction: float = DEFAULT_TEST_FRACTION
54
+ early_stopping_fraction: float = EARLY_STOPPING_FRACTION
55
  optimizer: str = "Adam"
56
  class_weighting: bool = True
57
  early_stopping_patience: int = 12
 
70
  raise ValueError("Learning rate must be in (0, 1].")
71
  if not 0 <= self.dropout < 1:
72
  raise ValueError("Dropout must be in [0, 1).")
73
+ if not 0.05 <= self.test_fraction <= 0.5:
74
+ raise ValueError("Test fraction must be between 0.05 and 0.5.")
75
+ if not 0.05 <= self.early_stopping_fraction <= 0.5:
76
+ raise ValueError("Early-stopping fraction must be between 0.05 and 0.5.")
77
  if self.optimizer.lower() not in {"adam", "sgd"}:
78
  raise ValueError("Optimizer must be Adam or SGD.")
79
  for name in ("biological_activation", "projection_activation", "fusion_activation"):
 
101
  @dataclass
102
  class TrainingResult:
103
  bundle: ModelBundle
104
+ test_predictions: pd.DataFrame
105
  confusion: np.ndarray
106
 
107
 
 
113
  torch.cuda.manual_seed_all(seed)
114
 
115
 
116
+ def stratified_split(
117
+ labels: np.ndarray, holdout_fraction: float, seed: int, holdout_name: str = "test"
118
  ) -> tuple[np.ndarray, np.ndarray]:
119
+ """Stratified (kept, held-out) index split shared by both omics and the labels."""
120
  classes, class_counts = np.unique(labels, return_counts=True)
121
  if class_counts.min() < 2:
122
  raise ValueError(
123
+ f"Each class needs at least two samples for a stratified {holdout_name} split."
124
  )
125
  sample_count = len(labels)
126
+ holdout_count = math.ceil(holdout_fraction * sample_count)
127
+ kept_count = sample_count - holdout_count
128
+ if min(holdout_count, kept_count) < len(classes):
129
  raise ValueError(
130
+ f"A {holdout_name} fraction of {holdout_fraction:.2f} on {sample_count} samples "
131
+ f"gives {kept_count} kept and {holdout_count} {holdout_name} samples; "
132
  f"each partition needs at least one sample per class ({len(classes)} classes) "
133
+ f"for a stratified split. Adjust the {holdout_name} fraction or add samples."
134
  )
135
 
136
  indices = np.arange(sample_count)
137
+ kept_index, holdout_index = train_test_split(
138
  indices,
139
+ test_size=holdout_fraction,
140
  random_state=seed,
141
  stratify=labels,
142
  )
143
+ return kept_index, holdout_index
144
+
145
+
146
+ def split_train_validation_test(
147
+ labels: np.ndarray, test_fraction: float, early_stopping_fraction: float, seed: int
148
+ ) -> tuple[np.ndarray, np.ndarray, np.ndarray]:
149
+ """Training, early-stopping validation and untouched test indices.
150
+
151
+ The test set is split off first; the validation set is then taken from the
152
+ training part only, so neither checkpoint selection nor fitting sees the test set.
153
+ """
154
+ development_index, test_index = stratified_split(labels, test_fraction, seed, "test")
155
+ fit_position, validation_position = stratified_split(
156
+ labels[development_index], early_stopping_fraction, seed, "early-stopping validation"
157
+ )
158
+ return development_index[fit_position], development_index[validation_position], test_index
159
 
160
 
161
  def _fit_scaler(values: np.ndarray) -> tuple[np.ndarray, np.ndarray]:
 
228
  hyperparameters.validate()
229
  set_reproducible_seed(hyperparameters.seed)
230
  labels = workspace.labels
231
+ train_index, validation_index, test_index = split_train_validation_test(
232
+ labels,
233
+ hyperparameters.test_fraction,
234
+ hyperparameters.early_stopping_fraction,
235
+ hyperparameters.seed,
236
  )
237
  gene_mean, gene_scale = _fit_scaler(
238
  workspace.gene_expression[train_index]
 
250
  dna_validation = _scale(
251
  workspace.dna_methylation[validation_index], dna_mean, dna_scale
252
  )
253
+ gene_test = _scale(workspace.gene_expression[test_index], gene_mean, gene_scale)
254
+ dna_test = _scale(workspace.dna_methylation[test_index], dna_mean, dna_scale)
255
  y_train = labels[train_index]
256
  y_validation = labels[validation_index]
257
+ y_test = labels[test_index]
258
 
259
  dataset = TensorDataset(
260
  torch.from_numpy(gene_train),
 
358
 
359
  if best_state is not None:
360
  model.load_state_dict(best_state)
361
+ # Scores come only from the untouched test set.
362
+ test_loss, probabilities, predictions = _evaluate(
363
  model,
364
+ gene_test,
365
+ dna_test,
366
+ y_test,
367
  device,
368
  )
369
  confusion = confusion_matrix(
370
+ y_test,
371
  predictions,
372
  labels=np.arange(len(workspace.label_names)),
373
  )
374
  metrics = {
375
+ "evaluation_split": "test",
376
+ "test_loss": test_loss,
377
+ "best_validation_loss": best_loss,
378
+ "accuracy": float(accuracy_score(y_test, predictions)),
379
  "f1_macro": float(
380
+ f1_score(y_test, predictions, average="macro", zero_division=0)
 
 
381
  ),
382
  "f1_weighted": float(
383
+ f1_score(y_test, predictions, average="weighted", zero_division=0)
 
 
384
  ),
385
  "precision_macro": float(
386
+ precision_score(y_test, predictions, average="macro", zero_division=0)
 
 
387
  ),
388
  "recall_macro": float(
389
+ recall_score(y_test, predictions, average="macro", zero_division=0)
 
 
390
  ),
391
  "epochs_completed": len(history),
392
  "training_samples": int(len(train_index)),
393
  "validation_samples": int(len(validation_index)),
394
+ "test_samples": int(len(test_index)),
395
  "device": str(device),
396
  }
397
  predictions_frame = pd.DataFrame(
398
  {
399
+ "sample_row": test_index,
400
+ "observed": [workspace.label_names[value] for value in y_test],
401
+ "predicted": [workspace.label_names[value] for value in predictions],
 
 
 
 
402
  }
403
  )
404
  for index, label in enumerate(workspace.label_names):
 
441
  )
442
  return TrainingResult(
443
  bundle=bundle,
444
+ test_predictions=predictions_frame,
445
  confusion=confusion,
446
  )
447
 
biolmnet/ui/components.py CHANGED
@@ -240,7 +240,7 @@ INTRO_CARDS = [
240
  "KEGG pathways, transcription-factor targets, protein interactions — so trained weights can be read "
241
  "back as pathways rather than as a black box."
242
  ),
243
- "This workbench runs the model end to end in five stages. On the bundled BRCA example it takes a few minutes.",
244
  "",
245
  ),
246
  (
@@ -280,7 +280,7 @@ INTRO_CARDS = [
280
  "Train Model",
281
  (
282
  "Hyperparameters sit on the left as a spec sheet. The run panel on the right reports <b>epoch, train "
283
- "and validation loss, accuracy and ETA</b> as the fit progresses."
284
  ),
285
  (
286
  "Controls lock for the duration of a run and the primary button reads <b>Training on ZeroGPU</b> "
@@ -308,7 +308,7 @@ INTRO_CARDS = [
308
  "Introduction · 06 / 06 · Stage 05",
309
  "Reading the results",
310
  (
311
- "Results collects everything a run produced: validation metrics, the training curve, the confusion "
312
  "matrix, an audit of how sparse each layer actually is, and the pathway attention weights — which is "
313
  "the part the architecture exists to give you."
314
  ),
 
240
  "KEGG pathways, transcription-factor targets, protein interactions — so trained weights can be read "
241
  "back as pathways rather than as a black box."
242
  ),
243
+ "This workbench runs the model end to end in five stages. On the bundled scTrioseq2 example it takes a few minutes.",
244
  "",
245
  ),
246
  (
 
280
  "Train Model",
281
  (
282
  "Hyperparameters sit on the left as a spec sheet. The run panel on the right reports <b>epoch, train "
283
+ "and validation loss, accuracy and ETA</b> as the fit progresses; final scores come from a held-out test set."
284
  ),
285
  (
286
  "Controls lock for the duration of a run and the primary button reads <b>Training on ZeroGPU</b> "
 
308
  "Introduction · 06 / 06 · Stage 05",
309
  "Reading the results",
310
  (
311
+ "Results collects everything a run produced: test-set metrics, the training curve, the confusion "
312
  "matrix, an audit of how sparse each layer actually is, and the pathway attention weights — which is "
313
  "the part the architecture exists to give you."
314
  ),
biolmnet/ui/styles.py CHANGED
@@ -315,6 +315,27 @@ body, .gradio-container {
315
  }
316
  .results-grid { padding: 26px 30px 0 !important; gap: 26px !important; align-items: flex-start !important; }
317
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
318
  .sect { margin: 0 0 9px; padding-bottom: 8px; border-bottom: 1px solid var(--color-text); }
319
  .sect-spaced { margin-top: 26px !important; }
320
 
 
315
  }
316
  .results-grid { padding: 26px 30px 0 !important; gap: 26px !important; align-items: flex-start !important; }
317
 
318
+ /* Results page reads as a report: white cards, body-font tables, full-strength text. */
319
+ #page-results .results-grid > .column,
320
+ #page-results .results-bottom {
321
+ background: #fff !important;
322
+ border: 1px solid var(--color-divider) !important;
323
+ border-radius: 6px;
324
+ padding: 16px 18px 18px !important;
325
+ }
326
+ #page-results .results-bottom { margin: 26px 30px 0 !important; }
327
+ #page-results .sect { font-size: 11.5px !important; color: var(--color-text) !important; letter-spacing: .08em; }
328
+ #page-results .stat { background: #fff; }
329
+ #page-results .stat .v { font-size: 34px; color: var(--color-text); }
330
+ #page-results .stat .k { font-size: 11px; color: var(--color-muted); }
331
+ /* Gradio draws visible rows as a virtual list of divs, not <td>s. */
332
+ #page-results table, #page-results table *, #page-results .virtual-body, #page-results .virtual-body * { font-family: var(--font-body) !important; font-size: 14px !important; font-variant-numeric: tabular-nums; }
333
+ #page-results thead th, #page-results thead th * { font: 600 11px/1.3 var(--font-mono) !important; letter-spacing: .06em; }
334
+ #page-results tbody td, #page-results .body-cell { color: var(--color-text) !important; background: #fff !important; }
335
+ #page-results tbody tr:nth-child(even) td, #page-results .virtual-row:nth-child(even) .body-cell { background: color-mix(in srgb, var(--color-accent) 5%, #fff) !important; }
336
+ #page-results thead th { font-size: 11px !important; color: var(--color-muted) !important; background: #fff !important; }
337
+ #page-results .table-wrap, #page-results .dataframe { background: #fff !important; }
338
+
339
  .sect { margin: 0 0 9px; padding-bottom: 8px; border-bottom: 1px solid var(--color-text); }
340
  .sect-spaced { margin-top: 26px !important; }
341
 
diagnostics/2026-10-30-other-datasets/results/configs_GBM.json ADDED
@@ -0,0 +1,59 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "D default (lab grid, tanh/relu/tanh, b32)": {},
3
+ "PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03)": {
4
+ "optimizer": "SGD",
5
+ "learning_rate": 0.01,
6
+ "weight_decay": 0.01,
7
+ "batch_size": 16,
8
+ "projection_dim": 64,
9
+ "epochs": 200,
10
+ "sgd_momentum": 0.03,
11
+ "biological_activation": "relu",
12
+ "projection_activation": "sigmoid",
13
+ "fusion_activation": "tanh"
14
+ },
15
+ "BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep)": {
16
+ "optimizer": "Adam",
17
+ "learning_rate": 0.001,
18
+ "weight_decay": 0.0001,
19
+ "batch_size": 16,
20
+ "projection_dim": 64,
21
+ "epochs": 100,
22
+ "biological_activation": "relu",
23
+ "projection_activation": "sigmoid",
24
+ "fusion_activation": "tanh"
25
+ },
26
+ "G1 grid b16 p64 tanh/relu/tanh": {
27
+ "learning_rate": 0.01,
28
+ "weight_decay": 0.01,
29
+ "epochs": 200,
30
+ "optimizer": "Adam",
31
+ "batch_size": 16,
32
+ "projection_dim": 64,
33
+ "biological_activation": "tanh",
34
+ "projection_activation": "relu",
35
+ "fusion_activation": "tanh"
36
+ },
37
+ "G2 grid b32 p64 tanh/relu/relu": {
38
+ "learning_rate": 0.01,
39
+ "weight_decay": 0.01,
40
+ "epochs": 200,
41
+ "optimizer": "Adam",
42
+ "batch_size": 32,
43
+ "projection_dim": 64,
44
+ "biological_activation": "tanh",
45
+ "projection_activation": "relu",
46
+ "fusion_activation": "relu"
47
+ },
48
+ "G3 grid b16 p128 tanh/tanh/tanh": {
49
+ "learning_rate": 0.01,
50
+ "weight_decay": 0.01,
51
+ "epochs": 200,
52
+ "optimizer": "Adam",
53
+ "batch_size": 16,
54
+ "projection_dim": 128,
55
+ "biological_activation": "tanh",
56
+ "projection_activation": "tanh",
57
+ "fusion_activation": "tanh"
58
+ }
59
+ }
diagnostics/2026-10-30-other-datasets/results/configs_common.json ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ {
2
+ "D default (lab grid, tanh/relu/tanh, b32)": {},
3
+ "PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03)": {"optimizer": "SGD", "learning_rate": 0.01, "weight_decay": 0.01, "batch_size": 16, "projection_dim": 64, "epochs": 200, "sgd_momentum": 0.03, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh"},
4
+ "BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep)": {"optimizer": "Adam", "learning_rate": 0.001, "weight_decay": 0.0001, "batch_size": 16, "projection_dim": 64, "epochs": 100, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh"}
5
+ }
diagnostics/2026-10-30-other-datasets/results/configs_scTrioseq2_top.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "S1 grid b32 p128 relu/tanh/tanh": {
3
+ "learning_rate": 0.01,
4
+ "weight_decay": 0.01,
5
+ "epochs": 200,
6
+ "optimizer": "Adam",
7
+ "batch_size": 32,
8
+ "projection_dim": 128,
9
+ "biological_activation": "relu",
10
+ "projection_activation": "tanh",
11
+ "fusion_activation": "tanh"
12
+ },
13
+ "S2 grid b32 p128 tanh/relu/tanh": {
14
+ "learning_rate": 0.01,
15
+ "weight_decay": 0.01,
16
+ "epochs": 200,
17
+ "optimizer": "Adam",
18
+ "batch_size": 32,
19
+ "projection_dim": 128,
20
+ "biological_activation": "tanh",
21
+ "projection_activation": "relu",
22
+ "fusion_activation": "tanh"
23
+ },
24
+ "S3 grid b32 p128 tanh/tanh/relu": {
25
+ "learning_rate": 0.01,
26
+ "weight_decay": 0.01,
27
+ "epochs": 200,
28
+ "optimizer": "Adam",
29
+ "batch_size": 32,
30
+ "projection_dim": 128,
31
+ "biological_activation": "tanh",
32
+ "projection_activation": "tanh",
33
+ "fusion_activation": "relu"
34
+ }
35
+ }
diagnostics/2026-10-30-other-datasets/results/grid_COAD_seed42.jsonl ADDED
@@ -0,0 +1,108 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "relu", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3909893035888672, "best_val_loss": 1.240048885345459}
2
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "relu", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.4272258877754211, "best_val_loss": 1.219422698020935}
3
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "relu", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3968784809112549, "best_val_loss": 1.2141286134719849}
4
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "relu", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 25, "best_epoch": 13, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3599404295285542, "best_val_loss": 1.2035648822784424}
5
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "sigmoid", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.1875, "f1_macro": 0.07894736842105263, "f1_weighted": 0.05921052631578947, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [0, 64, 0, 0], "classes_predicted": 1, "recall": [0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.4072861075401306, "best_val_loss": 1.2022572755813599}
6
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.4442652463912964, "best_val_loss": 1.1492609977722168}
7
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "sigmoid", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.4321313897768657, "best_val_loss": 1.1414717435836792}
8
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "tanh", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3983305891354878, "best_val_loss": 1.2026528120040894}
9
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "relu", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 29, "best_epoch": 17, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.384100337823232, "best_val_loss": 1.178083896636963}
10
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "tanh", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.4272340734799702, "best_val_loss": 1.2163499593734741}
11
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "tanh", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.396897792816162, "best_val_loss": 1.1978163719177246}
12
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "relu", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.379861573378245, "best_val_loss": 1.1927897930145264}
13
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "relu", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.1875, "f1_macro": 0.07894736842105263, "f1_weighted": 0.05921052631578947, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [0, 64, 0, 0], "classes_predicted": 1, "recall": [0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.3861074844996135, "best_val_loss": 1.159651517868042}
14
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "relu", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.4214906692504883, "best_val_loss": 1.1969072818756104}
15
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "sigmoid", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.412907252709071, "best_val_loss": 1.1631500720977783}
16
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "relu", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3805653254191081, "best_val_loss": 1.1614563465118408}
17
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.4119123816490173, "best_val_loss": 1.148013949394226}
18
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "sigmoid", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.1875, "f1_macro": 0.07894736842105263, "f1_weighted": 0.05921052631578947, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [0, 64, 0, 0], "classes_predicted": 1, "recall": [0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.3877576192220051, "best_val_loss": 1.1955989599227905}
19
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.4395293394724529, "best_val_loss": 1.1547168493270874}
20
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "sigmoid", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.43336021900177, "best_val_loss": 1.1464194059371948}
21
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "tanh", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.1875, "f1_macro": 0.07894736842105263, "f1_weighted": 0.05921052631578947, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [0, 64, 0, 0], "classes_predicted": 1, "recall": [0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.3968048294385274, "best_val_loss": 1.1957687139511108}
22
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "sigmoid", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 29, "best_epoch": 17, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.408077786366145, "best_val_loss": 1.1551365852355957}
23
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "tanh", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 22, "best_epoch": 10, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3777307669321697, "best_val_loss": 1.2420912981033325}
24
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "tanh", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3768468697865803, "best_val_loss": 1.2027405500411987}
25
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "tanh", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 23, "best_epoch": 11, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.4129614631334941, "best_val_loss": 1.1951897144317627}
26
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "relu", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.390831708908081, "best_val_loss": 1.2398327589035034}
27
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "relu", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.4272301991780598, "best_val_loss": 1.2192124128341675}
28
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "relu", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3094717065493267, "best_val_loss": 1.2129569053649902}
29
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "tanh", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 29, "best_epoch": 17, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3841036260128021, "best_val_loss": 1.1781153678894043}
30
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "sigmoid", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.1875, "f1_macro": 0.07894736842105263, "f1_weighted": 0.05921052631578947, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [0, 64, 0, 0], "classes_predicted": 1, "recall": [0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.4072534640630086, "best_val_loss": 1.2022759914398193}
31
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.444263478120168, "best_val_loss": 1.1492773294448853}
32
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "tanh", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 18, "best_epoch": 6, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.361961970726649, "best_val_loss": 1.2035990953445435}
33
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "sigmoid", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.432081123193105, "best_val_loss": 1.141425371170044}
34
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "tanh", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3989734649658203, "best_val_loss": 1.19907546043396}
35
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "relu", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3715976774692535, "best_val_loss": 1.2145028114318848}
36
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "tanh", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.4272460341453552, "best_val_loss": 1.2160921096801758}
37
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "tanh", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.1591469248135884, "best_val_loss": 1.197121024131775}
38
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "relu", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3704001903533936, "best_val_loss": 1.2720025777816772}
39
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "relu", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.377534846464793, "best_val_loss": 1.2321189641952515}
40
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "relu", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 29, "best_epoch": 17, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3864441911379497, "best_val_loss": 1.1905261278152466}
41
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "relu", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3983726104100545, "best_val_loss": 1.2272427082061768}
42
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "sigmoid", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3817058801651, "best_val_loss": 1.1318120956420898}
43
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.402257223924001, "best_val_loss": 1.192915916442871}
44
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "relu", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 18, "best_epoch": 6, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3573923707008362, "best_val_loss": 1.1654988527297974}
45
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "sigmoid", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.4438231190045674, "best_val_loss": 1.0954335927963257}
46
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "tanh", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3789408405621846, "best_val_loss": 1.2444725036621094}
47
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "sigmoid", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 15, "best_epoch": 3, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.36286727587382, "best_val_loss": 1.1704843044281006}
48
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "tanh", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3777499794960022, "best_val_loss": 1.2277497053146362}
49
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "tanh", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.1875, "f1_macro": 0.07894736842105263, "f1_weighted": 0.05921052631578947, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [0, 64, 0, 0], "classes_predicted": 1, "recall": [0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.3973684112230937, "best_val_loss": 1.2392377853393555}
50
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.4116776784261067, "best_val_loss": 1.1554261445999146}
51
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "relu", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3817211985588074, "best_val_loss": 1.2915568351745605}
52
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "relu", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.382285475730896, "best_val_loss": 1.2392115592956543}
53
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "sigmoid", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 20, "best_epoch": 8, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3634015421072643, "best_val_loss": 1.1400322914123535}
54
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "relu", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 18, "best_epoch": 6, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3922471006711323, "best_val_loss": 1.1967424154281616}
55
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "sigmoid", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3745754957199097, "best_val_loss": 1.1845200061798096}
56
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.4015939633051555, "best_val_loss": 1.2013922929763794}
57
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "tanh", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.1875, "f1_macro": 0.07894736842105263, "f1_weighted": 0.05921052631578947, "epochs_completed": 23, "best_epoch": 11, "pred_counts": [0, 64, 0, 0], "classes_predicted": 1, "recall": [0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.3559860289096832, "best_val_loss": 1.1589688062667847}
58
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "sigmoid", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.42834476629893, "best_val_loss": 1.0944057703018188}
59
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "tanh", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 16, "best_epoch": 4, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.38552329937617, "best_val_loss": 1.0078891515731812}
60
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "tanh", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 23, "best_epoch": 11, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3633587062358856, "best_val_loss": 1.1935319900512695}
61
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "tanh", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 23, "best_epoch": 11, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.366386592388153, "best_val_loss": 1.2155482769012451}
62
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "tanh", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 17, "best_epoch": 5, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.4204500516255696, "best_val_loss": 1.066306710243225}
63
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "tanh", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 18, "best_epoch": 6, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.4099347293376923, "best_val_loss": 1.1388736963272095}
64
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "relu", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3703823288281758, "best_val_loss": 1.272276520729065}
65
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "relu", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3775306940078735, "best_val_loss": 1.2322691679000854}
66
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "relu", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3062642415364583, "best_val_loss": 1.2243828773498535}
67
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "sigmoid", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3817062775293987, "best_val_loss": 1.1318864822387695}
68
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.4022567868232727, "best_val_loss": 1.1929216384887695}
69
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "sigmoid", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.4420830806096394, "best_val_loss": 1.0954800844192505}
70
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "tanh", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3793847759564717, "best_val_loss": 1.2465585470199585}
71
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "tanh", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3777526021003723, "best_val_loss": 1.2280129194259644}
72
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "relu", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.609375, "f1_macro": 0.19117647058823528, "f1_weighted": 0.5137867647058824, "epochs_completed": 52, "best_epoch": 40, "pred_counts": [59, 0, 5, 0], "classes_predicted": 2, "recall": [0.9069767441860465, 0.0, 0.0, 0.0], "final_train_loss": 1.21039746205012, "best_val_loss": 1.110617756843567}
73
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "tanh", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.421875, "f1_macro": 0.2357142857142857, "f1_weighted": 0.4147321428571428, "epochs_completed": 29, "best_epoch": 17, "pred_counts": [27, 2, 32, 3], "classes_predicted": 4, "recall": [0.4418604651162791, 0.0, 1.0, 0.0], "final_train_loss": 1.0241879522800446, "best_val_loss": 1.168595790863037}
74
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "relu", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 29, "best_epoch": 17, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3840636511643727, "best_val_loss": 1.1780965328216553}
75
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "relu", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.609375, "f1_macro": 0.18932038834951456, "f1_weighted": 0.5087985436893204, "epochs_completed": 36, "best_epoch": 24, "pred_counts": [60, 0, 2, 2], "classes_predicted": 3, "recall": [0.9069767441860465, 0.0, 0.0, 0.0], "final_train_loss": 1.0449495712916057, "best_val_loss": 1.1139811277389526}
76
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "sigmoid", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.413020024696986, "best_val_loss": 1.1631495952606201}
77
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.4119150936603546, "best_val_loss": 1.148013710975647}
78
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "sigmoid", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 29, "best_epoch": 17, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.408005913098653, "best_val_loss": 1.1541296243667603}
79
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "tanh", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3613113860289257, "best_val_loss": 1.2032543420791626}
80
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "tanh", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 29, "best_epoch": 17, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3841348787148793, "best_val_loss": 1.178093671798706}
81
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "tanh", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.625, "f1_macro": 0.19230769230769232, "f1_weighted": 0.5168269230769231, "epochs_completed": 29, "best_epoch": 17, "pred_counts": [61, 0, 3, 0], "classes_predicted": 2, "recall": [0.9302325581395349, 0.0, 0.0, 0.0], "final_train_loss": 1.00029818713665, "best_val_loss": 1.1232004165649414}
82
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "relu", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 18, "best_epoch": 6, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3678318361441295, "best_val_loss": 1.2158886194229126}
83
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "relu", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3622398177782695, "best_val_loss": 1.194149136543274}
84
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "relu", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 18, "best_epoch": 6, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.369064748287201, "best_val_loss": 1.203143835067749}
85
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "sigmoid", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 16, "best_epoch": 4, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3527783652146657, "best_val_loss": 1.2085145711898804}
86
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3967503309249878, "best_val_loss": 1.156850814819336}
87
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "sigmoid", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.125, "f1_macro": 0.05555555555555555, "f1_weighted": 0.027777777777777776, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [0, 0, 64, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0], "final_train_loss": 1.403890570004781, "best_val_loss": 1.1782804727554321}
88
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "tanh", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 20, "best_epoch": 8, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3543736040592194, "best_val_loss": 1.2200701236724854}
89
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "tanh", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3622565070788066, "best_val_loss": 1.1938904523849487}
90
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "tanh", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 18, "best_epoch": 6, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3698947727680206, "best_val_loss": 1.1976171731948853}
91
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "relu", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3639320333798726, "best_val_loss": 1.2280261516571045}
92
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "relu", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3910711805025737, "best_val_loss": 1.1997747421264648}
93
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "relu", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 23, "best_epoch": 11, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3795968095461528, "best_val_loss": 1.1490718126296997}
94
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "sigmoid", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 22, "best_epoch": 10, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3572869102160137, "best_val_loss": 1.1970808506011963}
95
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3977110385894775, "best_val_loss": 1.155841588973999}
96
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "sigmoid", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.125, "f1_macro": 0.05555555555555555, "f1_weighted": 0.027777777777777776, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [0, 0, 64, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0], "final_train_loss": 1.3980871538321178, "best_val_loss": 1.1731362342834473}
97
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "tanh", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 14, "best_epoch": 2, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3688914676507313, "best_val_loss": 1.0352522134780884}
98
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "tanh", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 16, "best_epoch": 4, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.380603164434433, "best_val_loss": 1.1692113876342773}
99
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "tanh", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 20, "best_epoch": 8, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3785443206628163, "best_val_loss": 1.1369270086288452}
100
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "relu", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 18, "best_epoch": 6, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3701550861199696, "best_val_loss": 1.2205222845077515}
101
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "relu", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3622499605019887, "best_val_loss": 1.1941701173782349}
102
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "relu", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.515625, "f1_macro": 0.19434269434269436, "f1_weighted": 0.48641636141636146, "epochs_completed": 54, "best_epoch": 42, "pred_counts": [48, 15, 1, 0], "classes_predicted": 3, "recall": [0.7441860465116279, 0.08333333333333333, 0.0, 0.0], "final_train_loss": 1.0227694809436798, "best_val_loss": 1.1140154600143433}
103
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "sigmoid", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 16, "best_epoch": 4, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3528986970583599, "best_val_loss": 1.2085161209106445}
104
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3967660466829936, "best_val_loss": 1.1568567752838135}
105
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "sigmoid", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.125, "f1_macro": 0.05555555555555555, "f1_weighted": 0.027777777777777776, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [0, 0, 64, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0], "final_train_loss": 1.4031570255756378, "best_val_loss": 1.1782838106155396}
106
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "tanh", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 20, "best_epoch": 8, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.353552758693695, "best_val_loss": 1.193418264389038}
107
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "tanh", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.671875, "f1_macro": 0.20093457943925233, "f1_weighted": 0.5400116822429907, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [64, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.3622585535049438, "best_val_loss": 1.1938930749893188}
108
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "tanh", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.421875, "f1_macro": 0.20723684210526316, "f1_weighted": 0.43585526315789475, "epochs_completed": 35, "best_epoch": 23, "pred_counts": [33, 28, 3, 0], "classes_predicted": 3, "recall": [0.5116279069767442, 0.4166666666666667, 0.0, 0.0], "final_train_loss": 1.0577431321144104, "best_val_loss": 1.0834901332855225}
diagnostics/2026-10-30-other-datasets/results/grid_COAD_shard0.log ADDED
@@ -0,0 +1,55 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [0] 0 {'projection_dim': 64, 'act1': 'relu', 'act2': 'relu', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
2
+ [0] 2 {'projection_dim': 64, 'act1': 'relu', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
3
+ [0] 4 {'projection_dim': 64, 'act1': 'relu', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
4
+ [0] 6 {'projection_dim': 64, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
5
+ [0] 8 {'projection_dim': 64, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
6
+ [0] 10 {'projection_dim': 64, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
7
+ [0] 12 {'projection_dim': 64, 'act1': 'relu', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
8
+ [0] 14 {'projection_dim': 64, 'act1': 'relu', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
9
+ [0] 16 {'projection_dim': 64, 'act1': 'relu', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
10
+ [0] 18 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
11
+ [0] 20 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
12
+ [0] 22 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
13
+ [0] 24 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
14
+ [0] 26 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
15
+ [0] 28 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
16
+ [0] 30 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
17
+ [0] 32 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
18
+ [0] 34 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
19
+ [0] 36 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'relu', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.191 classes=2
20
+ [0] 38 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
21
+ [0] 40 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.189 classes=3
22
+ [0] 42 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
23
+ [0] 44 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
24
+ [0] 46 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
25
+ [0] 48 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
26
+ [0] 50 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
27
+ [0] 52 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.192 classes=2
28
+ [0] 54 {'projection_dim': 128, 'act1': 'relu', 'act2': 'relu', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
29
+ [0] 56 {'projection_dim': 128, 'act1': 'relu', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
30
+ [0] 58 {'projection_dim': 128, 'act1': 'relu', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
31
+ [0] 60 {'projection_dim': 128, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
32
+ [0] 62 {'projection_dim': 128, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
33
+ [0] 64 {'projection_dim': 128, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.056 classes=1
34
+ [0] 66 {'projection_dim': 128, 'act1': 'relu', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
35
+ [0] 68 {'projection_dim': 128, 'act1': 'relu', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
36
+ [0] 70 {'projection_dim': 128, 'act1': 'relu', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
37
+ [0] 72 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
38
+ [0] 74 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
39
+ [0] 76 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
40
+ [0] 78 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
41
+ [0] 80 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
42
+ [0] 82 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.056 classes=1
43
+ [0] 84 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
44
+ [0] 86 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
45
+ [0] 88 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
46
+ [0] 90 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'relu', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
47
+ [0] 92 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
48
+ [0] 94 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.194 classes=3
49
+ [0] 96 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
50
+ [0] 98 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
51
+ [0] 100 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.056 classes=1
52
+ [0] 102 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
53
+ [0] 104 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
54
+ [0] 106 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.207 classes=3
55
+ SHARD 0 DONE
diagnostics/2026-10-30-other-datasets/results/grid_COAD_shard1.log ADDED
@@ -0,0 +1,55 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [1] 1 {'projection_dim': 64, 'act1': 'relu', 'act2': 'relu', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
2
+ [1] 3 {'projection_dim': 64, 'act1': 'relu', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
3
+ [1] 5 {'projection_dim': 64, 'act1': 'relu', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
4
+ [1] 7 {'projection_dim': 64, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
5
+ [1] 9 {'projection_dim': 64, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
6
+ [1] 11 {'projection_dim': 64, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
7
+ [1] 13 {'projection_dim': 64, 'act1': 'relu', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
8
+ [1] 15 {'projection_dim': 64, 'act1': 'relu', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
9
+ [1] 17 {'projection_dim': 64, 'act1': 'relu', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
10
+ [1] 19 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
11
+ [1] 21 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
12
+ [1] 23 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
13
+ [1] 25 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
14
+ [1] 27 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
15
+ [1] 29 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
16
+ [1] 31 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
17
+ [1] 33 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
18
+ [1] 35 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
19
+ [1] 37 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'relu', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
20
+ [1] 39 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
21
+ [1] 41 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
22
+ [1] 43 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
23
+ [1] 45 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
24
+ [1] 47 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
25
+ [1] 49 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
26
+ [1] 51 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
27
+ [1] 53 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
28
+ [1] 55 {'projection_dim': 128, 'act1': 'relu', 'act2': 'relu', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
29
+ [1] 57 {'projection_dim': 128, 'act1': 'relu', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
30
+ [1] 59 {'projection_dim': 128, 'act1': 'relu', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
31
+ [1] 61 {'projection_dim': 128, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
32
+ [1] 63 {'projection_dim': 128, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
33
+ [1] 65 {'projection_dim': 128, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
34
+ [1] 67 {'projection_dim': 128, 'act1': 'relu', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
35
+ [1] 69 {'projection_dim': 128, 'act1': 'relu', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
36
+ [1] 71 {'projection_dim': 128, 'act1': 'relu', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
37
+ [1] 73 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
38
+ [1] 75 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
39
+ [1] 77 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
40
+ [1] 79 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
41
+ [1] 81 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
42
+ [1] 83 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
43
+ [1] 85 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
44
+ [1] 87 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
45
+ [1] 89 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
46
+ [1] 91 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'relu', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
47
+ [1] 93 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
48
+ [1] 95 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
49
+ [1] 97 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
50
+ [1] 99 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
51
+ [1] 101 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
52
+ [1] 103 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
53
+ [1] 105 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.201 classes=1
54
+ [1] 107 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.236 classes=4
55
+ SHARD 1 DONE
diagnostics/2026-10-30-other-datasets/results/grid_COAD_summary.txt ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ configs: 108 collapsed (1 class): 102 all 5 classes predicted: 0 macro F1 >= 0.70: 0
2
+
3
+ TOP 20 by macro F1
4
+ optimizer batch_size projection_dim acts acc f1_macro f1_w min_recall best epochs pred
5
+ Adam 32 128 tanh/tanh/tanh 0.422 0.236 0.415 0.0 17 29 27/2/32/3
6
+ Adam 16 128 tanh/tanh/tanh 0.422 0.207 0.436 0.0 23 35 33/28/3/0
7
+ Adam 32 128 tanh/relu/sigmoid 0.672 0.201 0.540 0.0 1 13 64/0/0/0
8
+ Adam 16 64 tanh/sigmoid/sigmoid 0.672 0.201 0.540 0.0 1 13 64/0/0/0
9
+ Adam 16 64 tanh/sigmoid/relu 0.672 0.201 0.540 0.0 1 13 64/0/0/0
10
+ Adam 16 64 tanh/relu/sigmoid 0.672 0.201 0.540 0.0 17 29 64/0/0/0
11
+ Adam 32 128 tanh/tanh/sigmoid 0.672 0.201 0.540 0.0 1 13 64/0/0/0
12
+ Adam 32 128 tanh/tanh/relu 0.672 0.201 0.540 0.0 1 13 64/0/0/0
13
+ Adam 32 128 tanh/sigmoid/tanh 0.672 0.201 0.540 0.0 1 13 64/0/0/0
14
+ Adam 32 128 tanh/sigmoid/sigmoid 0.672 0.201 0.540 0.0 1 13 64/0/0/0
15
+ Adam 32 128 tanh/sigmoid/relu 0.672 0.201 0.540 0.0 1 13 64/0/0/0
16
+ Adam 32 128 tanh/relu/tanh 0.672 0.201 0.540 0.0 1 13 64/0/0/0
17
+ Adam 32 128 tanh/relu/relu 0.672 0.201 0.540 0.0 1 13 64/0/0/0
18
+ Adam 32 128 sigmoid/relu/relu 0.672 0.201 0.540 0.0 1 13 64/0/0/0
19
+ Adam 16 64 sigmoid/tanh/tanh 0.672 0.201 0.540 0.0 6 18 64/0/0/0
20
+ Adam 32 128 sigmoid/tanh/tanh 0.672 0.201 0.540 0.0 5 17 64/0/0/0
21
+ Adam 32 128 sigmoid/tanh/sigmoid 0.672 0.201 0.540 0.0 11 23 64/0/0/0
22
+ Adam 16 64 sigmoid/tanh/sigmoid 0.672 0.201 0.540 0.0 11 23 64/0/0/0
23
+ Adam 32 128 sigmoid/tanh/relu 0.672 0.201 0.540 0.0 4 16 64/0/0/0
24
+ Adam 32 128 sigmoid/sigmoid/tanh 0.672 0.201 0.540 0.0 1 13 64/0/0/0
25
+
26
+ BY optimizer
27
+ n f1_mean f1_max collapsed working
28
+ optimizer
29
+ Adam 108 0.189 0.236 102 0
30
+
31
+ BY batch_size
32
+ n f1_mean f1_max collapsed working
33
+ batch_size
34
+ 16 54 0.190 0.207 49 0
35
+ 32 54 0.188 0.236 53 0
36
+
37
+ BY projection_dim
38
+ n f1_mean f1_max collapsed working
39
+ projection_dim
40
+ 64 54 0.187 0.201 51 0
41
+ 128 54 0.191 0.236 51 0
42
+
43
+ BY act1
44
+ n f1_mean f1_max collapsed working
45
+ act1
46
+ relu 36 0.190 0.201 36 0
47
+ sigmoid 36 0.183 0.201 36 0
48
+ tanh 36 0.194 0.236 30 0
49
+
50
+ BY act2
51
+ n f1_mean f1_max collapsed working
52
+ act2
53
+ relu 36 0.197 0.201 33 0
54
+ sigmoid 36 0.179 0.201 36 0
55
+ tanh 36 0.192 0.236 33 0
56
+
57
+ BY act3
58
+ n f1_mean f1_max collapsed working
59
+ act3
60
+ relu 36 0.180 0.201 35 0
61
+ sigmoid 36 0.201 0.201 36 0
62
+ tanh 36 0.186 0.236 31 0
diagnostics/2026-10-30-other-datasets/results/grid_GBM_seed42.jsonl ADDED
@@ -0,0 +1,108 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "relu", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.32786885245901637, "f1_macro": 0.2549019607843137, "f1_weighted": 0.1984892317582771, "epochs_completed": 29, "best_epoch": 17, "pred_counts": [0, 4, 50, 7, 0], "classes_predicted": 3, "recall": [0.0, 0.0, 0.8333333333333334, 1.0, 0.0], "final_train_loss": 1.1435838182115816, "best_val_loss": 1.3830714225769043}
2
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "relu", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.3114754098360656, "f1_macro": 0.24315476190476187, "f1_weighted": 0.18264539422326306, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [49, 2, 0, 10, 0], "classes_predicted": 3, "recall": [0.8666666666666667, 0.08333333333333333, 0.0, 1.0, 0.0], "final_train_loss": 1.1442715349744579, "best_val_loss": 1.4128793478012085}
3
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "relu", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6144538994043902, "best_val_loss": 1.5881210565567017}
4
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "relu", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.3442622950819672, "f1_macro": 0.27715260017050297, "f1_weighted": 0.22553212304166145, "epochs_completed": 20, "best_epoch": 8, "pred_counts": [2, 1, 51, 7, 0], "classes_predicted": 4, "recall": [0.06666666666666667, 0.0, 0.8333333333333334, 1.0, 0.0], "final_train_loss": 1.134312358058867, "best_val_loss": 1.3640681505203247}
5
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "relu", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.620258975550125, "best_val_loss": 1.5694557428359985}
6
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "sigmoid", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6095422604044929, "best_val_loss": 1.5887278318405151}
7
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6183931729832635, "best_val_loss": 1.5824211835861206}
8
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "sigmoid", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 20, "best_epoch": 8, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.5866506738089472, "best_val_loss": 1.583168387413025}
9
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "relu", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.4098360655737705, "f1_macro": 0.4216529296327446, "f1_weighted": 0.3767104652391917, "epochs_completed": 38, "best_epoch": 26, "pred_counts": [10, 17, 23, 7, 4], "classes_predicted": 5, "recall": [0.2, 0.5, 0.5555555555555556, 1.0, 0.09090909090909091], "final_train_loss": 0.9489645453098693, "best_val_loss": 1.3042616844177246}
10
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "tanh", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.26229508196721313, "f1_macro": 0.2365079365079365, "f1_weighted": 0.1312776476710903, "epochs_completed": 26, "best_epoch": 14, "pred_counts": [0, 2, 0, 7, 52], "classes_predicted": 3, "recall": [0.0, 0.0, 0.0, 1.0, 1.0], "final_train_loss": 1.1652457596825772, "best_val_loss": 1.380925178527832}
11
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "tanh", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6144397434641102, "best_val_loss": 1.5881556272506714}
12
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "sigmoid", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 23, "best_epoch": 11, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6089167210573707, "best_val_loss": 1.5869073867797852}
13
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "tanh", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.4426229508196721, "f1_macro": 0.4243478260869565, "f1_weighted": 0.37208579018985294, "epochs_completed": 46, "best_epoch": 34, "pred_counts": [7, 11, 37, 6, 0], "classes_predicted": 4, "recall": [0.13333333333333333, 0.5, 0.7777777777777778, 1.0, 0.0], "final_train_loss": 0.9258286298298445, "best_val_loss": 1.3016177415847778}
14
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6170152627705225, "best_val_loss": 1.5710638761520386}
15
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "relu", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 41, "best_epoch": 29, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6094496145925887, "best_val_loss": 1.6034870147705078}
16
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "sigmoid", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6170247187379931, "best_val_loss": 1.5665905475616455}
17
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "relu", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6128497820734327, "best_val_loss": 1.5870773792266846}
18
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "relu", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 20, "best_epoch": 8, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.5926179729524206, "best_val_loss": 1.5838937759399414}
19
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "sigmoid", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.19672131147540983, "f1_macro": 0.06575342465753424, "f1_weighted": 0.06467549966314844, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [0, 61, 0, 0, 0], "classes_predicted": 1, "recall": [0.0, 1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6095491830117064, "best_val_loss": 1.5882856845855713}
20
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6180330619134538, "best_val_loss": 1.5822943449020386}
21
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "tanh", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.36065573770491804, "f1_macro": 0.35501116427432217, "f1_weighted": 0.2638512824535258, "epochs_completed": 38, "best_epoch": 26, "pred_counts": [42, 13, 0, 6, 0], "classes_predicted": 3, "recall": [0.7333333333333333, 0.5, 0.0, 1.0, 0.0], "final_train_loss": 0.9833631785840936, "best_val_loss": 1.3438191413879395}
22
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "sigmoid", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.19672131147540983, "f1_macro": 0.06575342465753424, "f1_weighted": 0.06467549966314844, "epochs_completed": 20, "best_epoch": 8, "pred_counts": [0, 61, 0, 0, 0], "classes_predicted": 1, "recall": [0.0, 1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.5987929196957031, "best_val_loss": 1.5732991695404053}
23
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "tanh", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.19672131147540983, "f1_macro": 0.06575342465753424, "f1_weighted": 0.06467549966314844, "epochs_completed": 17, "best_epoch": 5, "pred_counts": [0, 61, 0, 0, 0], "classes_predicted": 1, "recall": [0.0, 1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.609694270488343, "best_val_loss": 1.5873526334762573}
24
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "tanh", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6202207719042, "best_val_loss": 1.569453239440918}
25
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "tanh", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 14, "best_epoch": 2, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6109487528358002, "best_val_loss": 1.586789846420288}
26
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "tanh", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 20, "best_epoch": 8, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6045726257595208, "best_val_loss": 1.572192668914795}
27
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "relu", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.5573770491803278, "f1_macro": 0.6026491155902921, "f1_weighted": 0.5550033380313033, "epochs_completed": 55, "best_epoch": 43, "pred_counts": [15, 14, 16, 5, 11], "classes_predicted": 5, "recall": [0.4, 0.6666666666666666, 0.4444444444444444, 0.8, 0.7272727272727273], "final_train_loss": 0.7026482068124364, "best_val_loss": 0.9648270010948181}
28
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "relu", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6144514618024148, "best_val_loss": 1.5881175994873047}
29
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "tanh", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.47540983606557374, "f1_macro": 0.4936209736209737, "f1_weighted": 0.44901841623153105, "epochs_completed": 68, "best_epoch": 56, "pred_counts": [20, 14, 19, 6, 2], "classes_predicted": 5, "recall": [0.5333333333333333, 0.5, 0.5, 1.0, 0.09090909090909091], "final_train_loss": 0.925390218125015, "best_val_loss": 1.2859132289886475}
30
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "relu", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.5409836065573771, "f1_macro": 0.5952165331475676, "f1_weighted": 0.5327840302402201, "epochs_completed": 44, "best_epoch": 32, "pred_counts": [14, 15, 15, 6, 11], "classes_predicted": 5, "recall": [0.4, 0.6666666666666666, 0.3888888888888889, 1.0, 0.6363636363636364], "final_train_loss": 0.689169718919556, "best_val_loss": 0.9460645318031311}
31
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "sigmoid", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6095423255461814, "best_val_loss": 1.5887360572814941}
32
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "relu", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 31, "best_epoch": 19, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6098398719329001, "best_val_loss": 1.592193841934204}
33
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.618389784964056, "best_val_loss": 1.5824193954467773}
34
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "sigmoid", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 20, "best_epoch": 8, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.5867049440008696, "best_val_loss": 1.5831856727600098}
35
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "relu", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6179417176324813, "best_val_loss": 1.572983741760254}
36
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "relu", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 18, "best_epoch": 6, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6090717530641399, "best_val_loss": 1.587717056274414}
37
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "tanh", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.4426229508196721, "f1_macro": 0.5046972635207929, "f1_weighted": 0.4265845051765978, "epochs_completed": 54, "best_epoch": 42, "pred_counts": [9, 16, 16, 5, 15], "classes_predicted": 5, "recall": [0.13333333333333333, 0.8333333333333334, 0.3888888888888889, 1.0, 0.2727272727272727], "final_train_loss": 0.7952200685693918, "best_val_loss": 1.112304925918579}
38
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "tanh", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.614437150824917, "best_val_loss": 1.5881528854370117}
39
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "sigmoid", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 31, "best_epoch": 19, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6137747751559064, "best_val_loss": 1.5918093919754028}
40
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.616969569133279, "best_val_loss": 1.571669578552246}
41
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "tanh", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.5409836065573771, "f1_macro": 0.5611111111111111, "f1_weighted": 0.5231329690346084, "epochs_completed": 72, "best_epoch": 60, "pred_counts": [15, 12, 26, 4, 4], "classes_predicted": 5, "recall": [0.5333333333333333, 0.75, 0.6111111111111112, 0.8, 0.09090909090909091], "final_train_loss": 0.7009447650179836, "best_val_loss": 1.0161054134368896}
42
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "sigmoid", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 15, "best_epoch": 3, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6192130912197091, "best_val_loss": 1.558797001838684}
43
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "relu", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 22, "best_epoch": 10, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6107265740795864, "best_val_loss": 1.6001604795455933}
44
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "relu", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.619074611064515, "best_val_loss": 1.5918970108032227}
45
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "tanh", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 18, "best_epoch": 6, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6103808534601347, "best_val_loss": 1.5706861019134521}
46
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "tanh", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 15, "best_epoch": 3, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6180536890290473, "best_val_loss": 1.5744789838790894}
47
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "tanh", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 15, "best_epoch": 3, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6122698471194408, "best_val_loss": 1.5565847158432007}
48
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "relu", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.5245901639344263, "f1_macro": 0.5696894503346116, "f1_weighted": 0.528221029807492, "epochs_completed": 67, "best_epoch": 55, "pred_counts": [16, 10, 13, 7, 15], "classes_predicted": 5, "recall": [0.4, 0.6666666666666666, 0.4444444444444444, 1.0, 0.45454545454545453], "final_train_loss": 0.863623067980907, "best_val_loss": 1.244720458984375}
49
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "sigmoid", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 20, "best_epoch": 8, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6098492985866109, "best_val_loss": 1.5973776578903198}
50
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6153350964270003, "best_val_loss": 1.589383602142334}
51
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "sigmoid", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.616691382856317, "best_val_loss": 1.5708644390106201}
52
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "relu", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.4098360655737705, "f1_macro": 0.42869062720225515, "f1_weighted": 0.3598863203133123, "epochs_completed": 34, "best_epoch": 22, "pred_counts": [10, 21, 25, 5, 0], "classes_predicted": 4, "recall": [0.2, 0.6666666666666666, 0.5, 1.0, 0.0], "final_train_loss": 1.0046982973651157, "best_val_loss": 1.2295571565628052}
53
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "tanh", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6092947330631193, "best_val_loss": 1.5992294549942017}
54
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "relu", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6202599357386105, "best_val_loss": 1.5694561004638672}
55
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "tanh", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6190709957008154, "best_val_loss": 1.5918669700622559}
56
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "tanh", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.4262295081967213, "f1_macro": 0.47738951379925015, "f1_weighted": 0.41803807704283213, "epochs_completed": 63, "best_epoch": 51, "pred_counts": [9, 17, 16, 6, 13], "classes_predicted": 5, "recall": [0.26666666666666666, 0.5833333333333334, 0.3888888888888889, 1.0, 0.2727272727272727], "final_train_loss": 0.9748181830989858, "best_val_loss": 1.2949243783950806}
57
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "relu", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.5901639344262295, "f1_macro": 0.6113227081473926, "f1_weighted": 0.5856015283546151, "epochs_completed": 36, "best_epoch": 24, "pred_counts": [16, 13, 14, 6, 12], "classes_predicted": 5, "recall": [0.6, 0.6666666666666666, 0.4444444444444444, 0.8, 0.6363636363636364], "final_train_loss": 0.6505402593664784, "best_val_loss": 0.9335175156593323}
58
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "relu", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 22, "best_epoch": 10, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6097198496750795, "best_val_loss": 1.6002179384231567}
59
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "relu", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.618849476829904, "best_val_loss": 1.5916205644607544}
60
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "sigmoid", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.19672131147540983, "f1_macro": 0.06575342465753424, "f1_weighted": 0.06467549966314844, "epochs_completed": 28, "best_epoch": 16, "pred_counts": [0, 61, 0, 0, 0], "classes_predicted": 1, "recall": [0.0, 1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6088988670234472, "best_val_loss": 1.585577368736267}
61
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "relu", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 33, "best_epoch": 21, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6108017638732828, "best_val_loss": 1.5905486345291138}
62
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "sigmoid", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.19672131147540983, "f1_macro": 0.06575342465753424, "f1_weighted": 0.06467549966314844, "epochs_completed": 16, "best_epoch": 4, "pred_counts": [0, 61, 0, 0, 0], "classes_predicted": 1, "recall": [0.0, 1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6152916619035065, "best_val_loss": 1.5933729410171509}
63
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6170157793441104, "best_val_loss": 1.5710649490356445}
64
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.18032786885245902, "f1_macro": 0.061111111111111116, "f1_weighted": 0.055100182149362485, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [0, 0, 0, 0, 61], "classes_predicted": 1, "recall": [0.0, 0.0, 0.0, 0.0, 1.0], "final_train_loss": 1.6152529286556556, "best_val_loss": 1.589416265487671}
65
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "sigmoid", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6171117304452782, "best_val_loss": 1.5666223764419556}
66
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "sigmoid", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.19672131147540983, "f1_macro": 0.06575342465753424, "f1_weighted": 0.06467549966314844, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [0, 61, 0, 0, 0], "classes_predicted": 1, "recall": [0.0, 1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6170354990359863, "best_val_loss": 1.5747703313827515}
67
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "tanh", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.19672131147540983, "f1_macro": 0.06575342465753424, "f1_weighted": 0.06467549966314844, "epochs_completed": 34, "best_epoch": 22, "pred_counts": [0, 61, 0, 0, 0], "classes_predicted": 1, "recall": [0.0, 1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6191146816712259, "best_val_loss": 1.5948519706726074}
68
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "tanh", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.19672131147540983, "f1_macro": 0.06575342465753424, "f1_weighted": 0.06467549966314844, "epochs_completed": 14, "best_epoch": 2, "pred_counts": [0, 61, 0, 0, 0], "classes_predicted": 1, "recall": [0.0, 1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.618957109138614, "best_val_loss": 1.5863590240478516}
69
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "tanh", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.45901639344262296, "f1_macro": 0.45606060606060606, "f1_weighted": 0.40126676602086436, "epochs_completed": 34, "best_epoch": 22, "pred_counts": [9, 18, 26, 6, 2], "classes_predicted": 5, "recall": [0.2, 0.8333333333333334, 0.5555555555555556, 1.0, 0.0], "final_train_loss": 0.95225381492917, "best_val_loss": 1.1211503744125366}
70
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "tanh", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 32, "best_epoch": 20, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6117250710888638, "best_val_loss": 1.59309720993042}
71
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "tanh", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6202226225795642, "best_val_loss": 1.5694526433944702}
72
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "relu", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 22, "best_epoch": 10, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6107376058244967, "best_val_loss": 1.6002317667007446}
73
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "relu", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6190745810993383, "best_val_loss": 1.5918970108032227}
74
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "relu", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.4262295081967213, "f1_macro": 0.418008658008658, "f1_weighted": 0.36725569512454753, "epochs_completed": 38, "best_epoch": 26, "pred_counts": [5, 12, 38, 6, 0], "classes_predicted": 4, "recall": [0.2, 0.4166666666666667, 0.7222222222222222, 1.0, 0.0], "final_train_loss": 0.9166352361277804, "best_val_loss": 1.246969223022461}
75
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "tanh", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.5573770491803278, "f1_macro": 0.5724358974358974, "f1_weighted": 0.5215426649852879, "epochs_completed": 44, "best_epoch": 32, "pred_counts": [11, 14, 30, 5, 1], "classes_predicted": 5, "recall": [0.4, 0.75, 0.7222222222222222, 1.0, 0.09090909090909091], "final_train_loss": 0.8052770531242662, "best_val_loss": 1.033097505569458}
76
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "sigmoid", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 20, "best_epoch": 8, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6098517687594305, "best_val_loss": 1.597347617149353}
77
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6153320960008382, "best_val_loss": 1.5893851518630981}
78
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "sigmoid", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6170860131581624, "best_val_loss": 1.5712255239486694}
79
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "relu", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 31, "best_epoch": 19, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6098565710046904, "best_val_loss": 1.59080970287323}
80
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "tanh", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.45901639344262296, "f1_macro": 0.46420306513409954, "f1_weighted": 0.4178314804346461, "epochs_completed": 56, "best_epoch": 44, "pred_counts": [33, 13, 11, 4, 0], "classes_predicted": 4, "recall": [0.7333333333333333, 0.5833333333333334, 0.3333333333333333, 0.8, 0.0], "final_train_loss": 0.8745194908048286, "best_val_loss": 1.1733287572860718}
81
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "relu", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6190402755320399, "best_val_loss": 1.5791617631912231}
82
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "tanh", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6190676102872754, "best_val_loss": 1.5918607711791992}
83
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "tanh", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.36065573770491804, "f1_macro": 0.36431586914473135, "f1_weighted": 0.3001537186824034, "epochs_completed": 37, "best_epoch": 25, "pred_counts": [8, 18, 29, 6, 0], "classes_predicted": 4, "recall": [0.06666666666666667, 0.5, 0.5555555555555556, 1.0, 0.0], "final_train_loss": 0.8967835701228491, "best_val_loss": 1.2483084201812744}
84
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "relu", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.45901639344262296, "f1_macro": 0.4264264767490574, "f1_weighted": 0.38115612573571317, "epochs_completed": 44, "best_epoch": 32, "pred_counts": [5, 19, 31, 6, 0], "classes_predicted": 4, "recall": [0.13333333333333333, 0.5833333333333334, 0.7777777777777778, 1.0, 0.0], "final_train_loss": 1.0252954515277362, "best_val_loss": 1.3152095079421997}
85
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "sigmoid", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 22, "best_epoch": 10, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6170530039104607, "best_val_loss": 1.5765880346298218}
86
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 15, "best_epoch": 3, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6233703836065825, "best_val_loss": 1.588867425918579}
87
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "sigmoid", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.622798799165611, "best_val_loss": 1.5504661798477173}
88
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "tanh", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 31, "best_epoch": 19, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.60984157929655, "best_val_loss": 1.5915166139602661}
89
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "tanh", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6190970165481984, "best_val_loss": 1.579171061515808}
90
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "tanh", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.3442622950819672, "f1_macro": 0.24398699891657638, "f1_weighted": 0.1960463918441291, "epochs_completed": 17, "best_epoch": 5, "pred_counts": [0, 0, 53, 8, 0], "classes_predicted": 2, "recall": [0.0, 0.0, 0.8888888888888888, 1.0, 0.0], "final_train_loss": 1.1648317367001308, "best_val_loss": 1.3554661273956299}
91
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "relu", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 31, "best_epoch": 19, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6098419343187509, "best_val_loss": 1.5913604497909546}
92
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "relu", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 14, "best_epoch": 2, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6189456181448014, "best_val_loss": 1.5804989337921143}
93
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "relu", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 19, "best_epoch": 7, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6247041023494115, "best_val_loss": 1.5830711126327515}
94
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "sigmoid", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 26, "best_epoch": 14, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6229580614736172, "best_val_loss": 1.5805857181549072}
95
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 25, "best_epoch": 13, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6094001889880238, "best_val_loss": 1.5787322521209717}
96
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "sigmoid", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 35, "best_epoch": 23, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6275016473290698, "best_val_loss": 1.5733928680419922}
97
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "tanh", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 22, "best_epoch": 10, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6101267370369916, "best_val_loss": 1.5844577550888062}
98
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "tanh", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 23, "best_epoch": 11, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6240527063119607, "best_val_loss": 1.5801976919174194}
99
+ {"key": {"projection_dim": 128, "act1": "sigmoid", "act2": "tanh", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 14, "best_epoch": 2, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6243553265847794, "best_val_loss": 1.5606313943862915}
100
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "relu", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 15, "best_epoch": 3, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.609915028504335, "best_val_loss": 1.5926764011383057}
101
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "relu", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6190394078447519, "best_val_loss": 1.579162359237671}
102
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "relu", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.39344262295081966, "f1_macro": 0.4003809523809524, "f1_weighted": 0.32702576112412174, "epochs_completed": 46, "best_epoch": 34, "pred_counts": [6, 18, 32, 5, 0], "classes_predicted": 4, "recall": [0.06666666666666667, 0.5833333333333334, 0.6111111111111112, 1.0, 0.0], "final_train_loss": 0.9710713129877393, "best_val_loss": 1.2266485691070557}
103
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "sigmoid", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 22, "best_epoch": 10, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.610132048690254, "best_val_loss": 1.5811766386032104}
104
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 15, "best_epoch": 3, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.623384402749317, "best_val_loss": 1.5888473987579346}
105
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "sigmoid", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [61, 0, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0, 0.0], "final_train_loss": 1.6196753151429808, "best_val_loss": 1.5504440069198608}
106
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "tanh", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 31, "best_epoch": 19, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.60984196167826, "best_val_loss": 1.5914384126663208}
107
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "tanh", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 21, "best_epoch": 9, "pred_counts": [0, 0, 61, 0, 0], "classes_predicted": 1, "recall": [0.0, 0.0, 1.0, 0.0, 0.0], "final_train_loss": 1.6190990326834507, "best_val_loss": 1.579171895980835}
108
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "tanh", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.5901639344262295, "f1_macro": 0.5929791912732021, "f1_weighted": 0.5731392264342787, "epochs_completed": 78, "best_epoch": 66, "pred_counts": [14, 15, 17, 7, 8], "classes_predicted": 5, "recall": [0.5333333333333333, 0.8333333333333334, 0.6111111111111112, 1.0, 0.18181818181818182], "final_train_loss": 0.6872465102398981, "best_val_loss": 0.999639093875885}
diagnostics/2026-10-30-other-datasets/results/grid_GBM_shard0.log ADDED
@@ -0,0 +1,55 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [0] 0 {'projection_dim': 64, 'act1': 'relu', 'act2': 'relu', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.243 classes=3
2
+ [0] 2 {'projection_dim': 64, 'act1': 'relu', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
3
+ [0] 4 {'projection_dim': 64, 'act1': 'relu', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.422 classes=5
4
+ [0] 6 {'projection_dim': 64, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
5
+ [0] 8 {'projection_dim': 64, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
6
+ [0] 10 {'projection_dim': 64, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
7
+ [0] 12 {'projection_dim': 64, 'act1': 'relu', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.355 classes=3
8
+ [0] 14 {'projection_dim': 64, 'act1': 'relu', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
9
+ [0] 16 {'projection_dim': 64, 'act1': 'relu', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.494 classes=5
10
+ [0] 18 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
11
+ [0] 20 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
12
+ [0] 22 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
13
+ [0] 24 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
14
+ [0] 26 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
15
+ [0] 28 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
16
+ [0] 30 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
17
+ [0] 32 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
18
+ [0] 34 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
19
+ [0] 36 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'relu', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.429 classes=4
20
+ [0] 38 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
21
+ [0] 40 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.611 classes=5
22
+ [0] 42 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.066 classes=1
23
+ [0] 44 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
24
+ [0] 46 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
25
+ [0] 48 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.456 classes=5
26
+ [0] 50 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
27
+ [0] 52 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.572 classes=5
28
+ [0] 54 {'projection_dim': 128, 'act1': 'relu', 'act2': 'relu', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
29
+ [0] 56 {'projection_dim': 128, 'act1': 'relu', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
30
+ [0] 58 {'projection_dim': 128, 'act1': 'relu', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.426 classes=4
31
+ [0] 60 {'projection_dim': 128, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
32
+ [0] 62 {'projection_dim': 128, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
33
+ [0] 64 {'projection_dim': 128, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
34
+ [0] 66 {'projection_dim': 128, 'act1': 'relu', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
35
+ [0] 68 {'projection_dim': 128, 'act1': 'relu', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
36
+ [0] 70 {'projection_dim': 128, 'act1': 'relu', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.244 classes=2
37
+ [0] 72 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
38
+ [0] 74 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
39
+ [0] 76 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
40
+ [0] 78 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
41
+ [0] 80 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
42
+ [0] 82 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
43
+ [0] 84 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
44
+ [0] 86 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
45
+ [0] 88 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
46
+ [0] 90 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'relu', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
47
+ [0] 92 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
48
+ [0] 94 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.400 classes=4
49
+ [0] 96 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
50
+ [0] 98 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
51
+ [0] 100 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
52
+ [0] 102 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
53
+ [0] 104 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
54
+ [0] 106 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.593 classes=5
55
+ SHARD 0 DONE
diagnostics/2026-10-30-other-datasets/results/grid_GBM_shard1.log ADDED
@@ -0,0 +1,55 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [1] 1 {'projection_dim': 64, 'act1': 'relu', 'act2': 'relu', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.255 classes=3
2
+ [1] 3 {'projection_dim': 64, 'act1': 'relu', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
3
+ [1] 5 {'projection_dim': 64, 'act1': 'relu', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.277 classes=4
4
+ [1] 7 {'projection_dim': 64, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
5
+ [1] 9 {'projection_dim': 64, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
6
+ [1] 11 {'projection_dim': 64, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
7
+ [1] 13 {'projection_dim': 64, 'act1': 'relu', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.237 classes=3
8
+ [1] 15 {'projection_dim': 64, 'act1': 'relu', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
9
+ [1] 17 {'projection_dim': 64, 'act1': 'relu', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.424 classes=4
10
+ [1] 19 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
11
+ [1] 21 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
12
+ [1] 23 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
13
+ [1] 25 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.066 classes=1
14
+ [1] 27 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
15
+ [1] 29 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.066 classes=1
16
+ [1] 31 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.066 classes=1
17
+ [1] 33 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
18
+ [1] 35 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
19
+ [1] 37 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'relu', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.603 classes=5
20
+ [1] 39 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
21
+ [1] 41 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.595 classes=5
22
+ [1] 43 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
23
+ [1] 45 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
24
+ [1] 47 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
25
+ [1] 49 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.505 classes=5
26
+ [1] 51 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
27
+ [1] 53 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.561 classes=5
28
+ [1] 55 {'projection_dim': 128, 'act1': 'relu', 'act2': 'relu', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
29
+ [1] 57 {'projection_dim': 128, 'act1': 'relu', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
30
+ [1] 59 {'projection_dim': 128, 'act1': 'relu', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.570 classes=5
31
+ [1] 61 {'projection_dim': 128, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
32
+ [1] 63 {'projection_dim': 128, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
33
+ [1] 65 {'projection_dim': 128, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
34
+ [1] 67 {'projection_dim': 128, 'act1': 'relu', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
35
+ [1] 69 {'projection_dim': 128, 'act1': 'relu', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
36
+ [1] 71 {'projection_dim': 128, 'act1': 'relu', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.477 classes=5
37
+ [1] 73 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
38
+ [1] 75 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
39
+ [1] 77 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
40
+ [1] 79 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.066 classes=1
41
+ [1] 81 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.061 classes=1
42
+ [1] 83 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.066 classes=1
43
+ [1] 85 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.066 classes=1
44
+ [1] 87 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.066 classes=1
45
+ [1] 89 {'projection_dim': 128, 'act1': 'sigmoid', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
46
+ [1] 91 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'relu', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
47
+ [1] 93 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
48
+ [1] 95 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.418 classes=4
49
+ [1] 97 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
50
+ [1] 99 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
51
+ [1] 101 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.091 classes=1
52
+ [1] 103 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.464 classes=4
53
+ [1] 105 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.079 classes=1
54
+ [1] 107 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.364 classes=4
55
+ SHARD 1 DONE
diagnostics/2026-10-30-other-datasets/results/grid_GBM_summary.txt ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ configs: 108 collapsed (1 class): 83 all 5 classes predicted: 12 macro F1 >= 0.70: 0
2
+
3
+ TOP 20 by macro F1
4
+ optimizer batch_size projection_dim acts acc f1_macro f1_w min_recall best epochs pred
5
+ Adam 16 64 tanh/relu/tanh 0.590 0.611 0.586 0.444 24 36 16/13/14/6/12
6
+ Adam 32 64 tanh/relu/relu 0.557 0.603 0.555 0.400 43 55 15/14/16/5/11
7
+ Adam 32 64 tanh/relu/tanh 0.541 0.595 0.533 0.389 32 44 14/15/15/6/11
8
+ Adam 16 128 tanh/tanh/tanh 0.590 0.593 0.573 0.182 66 78 14/15/17/7/8
9
+ Adam 16 64 tanh/tanh/tanh 0.557 0.572 0.522 0.091 32 44 11/14/30/5/1
10
+ Adam 32 128 relu/relu/tanh 0.525 0.570 0.528 0.400 55 67 16/10/13/7/15
11
+ Adam 32 64 tanh/tanh/tanh 0.541 0.561 0.523 0.091 60 72 15/12/26/4/4
12
+ Adam 32 64 tanh/tanh/relu 0.443 0.505 0.427 0.133 42 54 9/16/16/5/15
13
+ Adam 16 64 relu/tanh/tanh 0.475 0.494 0.449 0.091 56 68 20/14/19/6/2
14
+ Adam 32 128 relu/tanh/tanh 0.426 0.477 0.418 0.267 51 63 9/17/16/6/13
15
+ Adam 32 128 tanh/tanh/relu 0.459 0.464 0.418 0.000 44 56 33/13/11/4/0
16
+ Adam 16 64 tanh/tanh/relu 0.459 0.456 0.401 0.000 22 34 9/18/26/6/2
17
+ Adam 16 64 tanh/relu/relu 0.410 0.429 0.360 0.000 22 34 10/21/25/5/0
18
+ Adam 16 128 relu/relu/tanh 0.459 0.426 0.381 0.000 32 44 5/19/31/6/0
19
+ Adam 32 64 relu/tanh/tanh 0.443 0.424 0.372 0.000 34 46 7/11/37/6/0
20
+ Adam 16 64 relu/relu/tanh 0.410 0.422 0.377 0.091 26 38 10/17/23/7/4
21
+ Adam 32 128 tanh/relu/tanh 0.426 0.418 0.367 0.000 26 38 5/12/38/6/0
22
+ Adam 16 128 tanh/relu/tanh 0.393 0.400 0.327 0.000 34 46 6/18/32/5/0
23
+ Adam 32 128 tanh/tanh/tanh 0.361 0.364 0.300 0.000 25 37 8/18/29/6/0
24
+ Adam 16 64 relu/tanh/relu 0.361 0.355 0.264 0.000 26 38 42/13/0/6/0
25
+
26
+ BY optimizer
27
+ n f1_mean f1_max collapsed working
28
+ optimizer
29
+ Adam 108 0.166 0.611 83 0
30
+
31
+ BY batch_size
32
+ n f1_mean f1_max collapsed working
33
+ batch_size
34
+ 16 54 0.165 0.611 42 0
35
+ 32 54 0.168 0.603 41 0
36
+
37
+ BY projection_dim
38
+ n f1_mean f1_max collapsed working
39
+ projection_dim
40
+ 64 54 0.189 0.611 38 0
41
+ 128 54 0.144 0.593 45 0
42
+
43
+ BY act1
44
+ n f1_mean f1_max collapsed working
45
+ act1
46
+ relu 36 0.180 0.570 24 0
47
+ sigmoid 36 0.082 0.091 36 0
48
+ tanh 36 0.237 0.611 23 0
49
+
50
+ BY act2
51
+ n f1_mean f1_max collapsed working
52
+ act2
53
+ relu 36 0.204 0.611 24 0
54
+ sigmoid 36 0.082 0.091 36 0
55
+ tanh 36 0.213 0.593 23 0
56
+
57
+ BY act3
58
+ n f1_mean f1_max collapsed working
59
+ act3
60
+ relu 36 0.161 0.603 27 0
61
+ sigmoid 36 0.085 0.091 36 0
62
+ tanh 36 0.254 0.611 20 0
diagnostics/2026-10-30-other-datasets/results/grid_scTrioseq2_reduced_shard0.log ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ [0] 8 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'relu', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.758 classes=4
2
+ [0] 12 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.767 classes=4
3
+ [0] 16 {'projection_dim': 128, 'act1': 'relu', 'act2': 'relu', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.579 classes=3
4
+ [0] 20 {'projection_dim': 128, 'act1': 'relu', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.698 classes=4
5
+ [0] 24 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'relu', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.540 classes=3
diagnostics/2026-10-30-other-datasets/results/grid_scTrioseq2_reduced_shard1.log ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ [1] 9 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'relu', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.785 classes=4
2
+ [1] 13 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.752 classes=4
3
+ [1] 17 {'projection_dim': 128, 'act1': 'relu', 'act2': 'relu', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.681 classes=4
4
+ [1] 21 {'projection_dim': 128, 'act1': 'relu', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.714 classes=4
5
+ [1] 25 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'relu', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.790 classes=4
6
+ [1] 29 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.793 classes=4
7
+ SHARD 1 DONE
diagnostics/2026-10-30-other-datasets/results/grid_scTrioseq2_reduced_shard2.log ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ [2] 6 {'projection_dim': 64, 'act1': 'relu', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.766 classes=4
2
+ [2] 10 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.782 classes=4
3
+ [2] 14 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.626 classes=4
4
+ [2] 18 {'projection_dim': 128, 'act1': 'relu', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.773 classes=4
5
+ [2] 22 {'projection_dim': 128, 'act1': 'relu', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.770 classes=4
diagnostics/2026-10-30-other-datasets/results/grid_scTrioseq2_reduced_shard3.log ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ [3] 11 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.839 classes=4
2
+ [3] 15 {'projection_dim': 64, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.667 classes=4
3
+ [3] 19 {'projection_dim': 128, 'act1': 'relu', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.744 classes=4
4
+ [3] 23 {'projection_dim': 128, 'act1': 'relu', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.803 classes=4
5
+ [3] 27 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.793 classes=4
6
+ [3] 31 {'projection_dim': 128, 'act1': 'tanh', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.737 classes=4
7
+ SHARD 3 DONE
diagnostics/2026-10-30-other-datasets/results/grid_scTrioseq2_seed42.jsonl ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "relu", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 30, "best_epoch": 18, "pred_counts": [0, 0, 0, 153], "classes_predicted": 1, "recall": [0.0, 0.0, 0.0, 1.0], "final_train_loss": 1.3873135102422613, "best_val_loss": 1.3700900077819824}
2
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "relu", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 25, "best_epoch": 13, "pred_counts": [0, 0, 0, 153], "classes_predicted": 1, "recall": [0.0, 0.0, 0.0, 1.0], "final_train_loss": 1.3934418590445268, "best_val_loss": 1.3605917692184448}
3
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "relu", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.7058823529411765, "f1_macro": 0.721741277442332, "f1_weighted": 0.7041269293147558, "epochs_completed": 44, "best_epoch": 32, "pred_counts": [33, 44, 20, 56], "classes_predicted": 4, "recall": [0.5789473684210527, 0.896551724137931, 0.85, 0.6515151515151515], "final_train_loss": 0.6618062205481947, "best_val_loss": 0.7239845395088196}
4
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "relu", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.5098039215686274, "f1_macro": 0.5525734992198408, "f1_weighted": 0.4410209486329365, "epochs_completed": 47, "best_epoch": 35, "pred_counts": [70, 43, 24, 16], "classes_predicted": 4, "recall": [0.7631578947368421, 0.7931034482758621, 0.95, 0.10606060606060606], "final_train_loss": 0.6456542877774489, "best_val_loss": 0.7737857103347778}
5
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "sigmoid", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [0, 0, 0, 153], "classes_predicted": 1, "recall": [0.0, 0.0, 0.0, 1.0], "final_train_loss": 1.3922991648054959, "best_val_loss": 1.3530312776565552}
6
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "relu", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 19, "best_epoch": 7, "pred_counts": [0, 0, 0, 153], "classes_predicted": 1, "recall": [0.0, 0.0, 0.0, 1.0], "final_train_loss": 1.3946980781722487, "best_val_loss": 1.3498953580856323}
7
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 19, "best_epoch": 7, "pred_counts": [0, 0, 0, 153], "classes_predicted": 1, "recall": [0.0, 0.0, 0.0, 1.0], "final_train_loss": 1.3929637231324847, "best_val_loss": 1.359687089920044}
8
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "sigmoid", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 24, "best_epoch": 12, "pred_counts": [0, 0, 0, 153], "classes_predicted": 1, "recall": [0.0, 0.0, 0.0, 1.0], "final_train_loss": 1.391294981303968, "best_val_loss": 1.345849871635437}
9
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "relu", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.5947712418300654, "f1_macro": 0.549165120593692, "f1_weighted": 0.5109437714479731, "epochs_completed": 32, "best_epoch": 20, "pred_counts": [0, 48, 24, 81], "classes_predicted": 3, "recall": [0.0, 0.896551724137931, 1.0, 0.6818181818181818], "final_train_loss": 0.7018749588414243, "best_val_loss": 0.7821705937385559}
10
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "tanh", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 30, "best_epoch": 18, "pred_counts": [0, 0, 0, 153], "classes_predicted": 1, "recall": [0.0, 0.0, 0.0, 1.0], "final_train_loss": 1.3873152941988225, "best_val_loss": 1.370130181312561}
11
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "tanh", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 25, "best_epoch": 13, "pred_counts": [0, 0, 0, 153], "classes_predicted": 1, "recall": [0.0, 0.0, 0.0, 1.0], "final_train_loss": 1.3934426830525983, "best_val_loss": 1.3605964183807373}
12
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "sigmoid", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.24836601307189543, "f1_macro": 0.09947643979057591, "f1_weighted": 0.09882626698148718, "epochs_completed": 23, "best_epoch": 11, "pred_counts": [153, 0, 0, 0], "classes_predicted": 1, "recall": [1.0, 0.0, 0.0, 0.0], "final_train_loss": 1.392561299759045, "best_val_loss": 1.347267746925354}
13
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 28, "best_epoch": 16, "pred_counts": [0, 0, 0, 153], "classes_predicted": 1, "recall": [0.0, 0.0, 0.0, 1.0], "final_train_loss": 1.3941203732239573, "best_val_loss": 1.3516309261322021}
14
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "tanh", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.7320261437908496, "f1_macro": 0.7494775624212588, "f1_weighted": 0.7316650214051035, "epochs_completed": 58, "best_epoch": 46, "pred_counts": [38, 41, 23, 51], "classes_predicted": 4, "recall": [0.7105263157894737, 0.8275862068965517, 0.95, 0.6363636363636364], "final_train_loss": 0.5055314668437891, "best_val_loss": 0.593417227268219}
15
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "relu", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 14, "best_epoch": 2, "pred_counts": [0, 0, 0, 153], "classes_predicted": 1, "recall": [0.0, 0.0, 0.0, 1.0], "final_train_loss": 1.3876600495555944, "best_val_loss": 1.3639287948608398}
16
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "sigmoid", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [0, 0, 0, 153], "classes_predicted": 1, "recall": [0.0, 0.0, 0.0, 1.0], "final_train_loss": 1.3996041218439739, "best_val_loss": 1.3484610319137573}
17
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "relu", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 25, "best_epoch": 13, "pred_counts": [0, 0, 0, 153], "classes_predicted": 1, "recall": [0.0, 0.0, 0.0, 1.0], "final_train_loss": 1.3939158916473389, "best_val_loss": 1.3590943813323975}
18
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "relu", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [0, 0, 0, 153], "classes_predicted": 1, "recall": [0.0, 0.0, 0.0, 1.0], "final_train_loss": 1.3873379209585357, "best_val_loss": 1.3682854175567627}
19
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "sigmoid", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 13, "best_epoch": 1, "pred_counts": [0, 0, 0, 153], "classes_predicted": 1, "recall": [0.0, 0.0, 0.0, 1.0], "final_train_loss": 1.393721109942386, "best_val_loss": 1.3561733961105347}
20
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "tanh", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.6993464052287581, "f1_macro": 0.7149999077501911, "f1_weighted": 0.6956752523035177, "epochs_completed": 32, "best_epoch": 20, "pred_counts": [33, 38, 21, 61], "classes_predicted": 4, "recall": [0.5526315789473685, 0.8620689655172413, 0.85, 0.6666666666666666], "final_train_loss": 0.679388257495144, "best_val_loss": 0.7309417724609375}
21
+ {"key": {"projection_dim": 64, "act1": "sigmoid", "act2": "sigmoid", "act3": "sigmoid", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "sigmoid", "projection_activation": "sigmoid", "fusion_activation": "sigmoid", "sgd_momentum": 0.0}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 19, "best_epoch": 7, "pred_counts": [0, 0, 0, 153], "classes_predicted": 1, "recall": [0.0, 0.0, 0.0, 1.0], "final_train_loss": 1.3931162650125068, "best_val_loss": 1.3597816228866577}
22
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "relu", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.7647058823529411, "f1_macro": 0.7852310106974286, "f1_weighted": 0.762519747023698, "epochs_completed": 79, "best_epoch": 67, "pred_counts": [42, 38, 22, 51], "classes_predicted": 4, "recall": [0.7631578947368421, 0.896551724137931, 0.95, 0.6515151515151515], "final_train_loss": 0.45690638506621645, "best_val_loss": 0.5384410619735718}
23
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "relu", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.8300653594771242, "f1_macro": 0.8389094286870471, "f1_weighted": 0.8280242914220465, "epochs_completed": 87, "best_epoch": 75, "pred_counts": [30, 36, 21, 66], "classes_predicted": 4, "recall": [0.6842105263157895, 0.9310344827586207, 0.95, 0.8333333333333334], "final_train_loss": 0.4291579744272065, "best_val_loss": 0.48823636770248413}
24
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "tanh", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.7450980392156863, "f1_macro": 0.7523300374306969, "f1_weighted": 0.7454765748454624, "epochs_completed": 34, "best_epoch": 22, "pred_counts": [29, 45, 16, 63], "classes_predicted": 4, "recall": [0.6052631578947368, 0.9310344827586207, 0.75, 0.7424242424242424], "final_train_loss": 0.5726166189762584, "best_val_loss": 0.6654586791992188}
25
+ {"key": {"projection_dim": 64, "act1": "relu", "act2": "tanh", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.7450980392156863, "f1_macro": 0.765710795935254, "f1_weighted": 0.7464899215976736, "epochs_completed": 62, "best_epoch": 50, "pred_counts": [47, 43, 18, 45], "classes_predicted": 4, "recall": [0.8421052631578947, 0.8620689655172413, 0.85, 0.6060606060606061], "final_train_loss": 0.5049633603346976, "best_val_loss": 0.6015363931655884}
26
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "tanh", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.6143790849673203, "f1_macro": 0.6668107880818652, "f1_weighted": 0.589617804009155, "epochs_completed": 53, "best_epoch": 41, "pred_counts": [66, 38, 21, 28], "classes_predicted": 4, "recall": [0.7631578947368421, 0.9310344827586207, 0.9, 0.30303030303030304], "final_train_loss": 0.57662471919729, "best_val_loss": 0.697319746017456}
27
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "relu", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.7516339869281046, "f1_macro": 0.7584784877811492, "f1_weighted": 0.751614431101028, "epochs_completed": 78, "best_epoch": 66, "pred_counts": [34, 38, 23, 58], "classes_predicted": 4, "recall": [0.6842105263157895, 0.8275862068965517, 0.9, 0.7121212121212122], "final_train_loss": 0.49395094419780533, "best_val_loss": 0.5397946238517761}
28
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "relu", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.6535947712418301, "f1_macro": 0.6810335457084683, "f1_weighted": 0.647391953704101, "epochs_completed": 37, "best_epoch": 25, "pred_counts": [36, 46, 18, 53], "classes_predicted": 4, "recall": [0.5, 0.9655172413793104, 0.8, 0.5606060606060606], "final_train_loss": 0.7002025215249312, "best_val_loss": 0.7461352348327637}
29
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "relu", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.7320261437908496, "f1_macro": 0.7436367558239863, "f1_weighted": 0.7365263045108754, "epochs_completed": 49, "best_epoch": 37, "pred_counts": [38, 43, 16, 56], "classes_predicted": 4, "recall": [0.7368421052631579, 0.8275862068965517, 0.75, 0.6818181818181818], "final_train_loss": 0.4792041349829289, "best_val_loss": 0.5984147191047668}
30
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "tanh", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.7581699346405228, "f1_macro": 0.7671443260981062, "f1_weighted": 0.7502624803508725, "epochs_completed": 31, "best_epoch": 19, "pred_counts": [22, 44, 22, 65], "classes_predicted": 4, "recall": [0.5, 0.9655172413793104, 0.95, 0.7575757575757576], "final_train_loss": 0.6614926405120314, "best_val_loss": 0.6519563794136047}
31
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "tanh", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.6993464052287581, "f1_macro": 0.7140984015984017, "f1_weighted": 0.6961626608685433, "epochs_completed": 58, "best_epoch": 46, "pred_counts": [32, 37, 20, 64], "classes_predicted": 4, "recall": [0.5263157894736842, 0.8275862068965517, 0.85, 0.696969696969697], "final_train_loss": 0.5954567716832746, "best_val_loss": 0.6591925621032715}
32
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "relu", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.7843137254901961, "f1_macro": 0.7824440052700923, "f1_weighted": 0.7863976508477787, "epochs_completed": 58, "best_epoch": 46, "pred_counts": [31, 40, 16, 66], "classes_predicted": 4, "recall": [0.6842105263157895, 0.8620689655172413, 0.75, 0.8181818181818182], "final_train_loss": 0.48087429320603087, "best_val_loss": 0.5209798216819763}
33
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "tanh", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.7973856209150327, "f1_macro": 0.8029562542720436, "f1_weighted": 0.8037893307243152, "epochs_completed": 62, "best_epoch": 50, "pred_counts": [38, 48, 18, 49], "classes_predicted": 4, "recall": [0.868421052631579, 0.9310344827586207, 0.8, 0.696969696969697], "final_train_loss": 0.4686367522206223, "best_val_loss": 0.5512884259223938}
34
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "relu", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.7777777777777778, "f1_macro": 0.7896159238488785, "f1_weighted": 0.7760403403785758, "epochs_completed": 62, "best_epoch": 50, "pred_counts": [39, 35, 24, 55], "classes_predicted": 4, "recall": [0.7631578947368421, 0.8620689655172413, 0.95, 0.696969696969697], "final_train_loss": 0.5012885635359245, "best_val_loss": 0.5802960991859436}
35
+ {"key": {"projection_dim": 64, "act1": "tanh", "act2": "tanh", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.5751633986928104, "f1_macro": 0.6264729173265758, "f1_weighted": 0.5401471157568719, "epochs_completed": 38, "best_epoch": 26, "pred_counts": [74, 36, 21, 22], "classes_predicted": 4, "recall": [0.7894736842105263, 0.896551724137931, 0.85, 0.22727272727272727], "final_train_loss": 0.6228602784767485, "best_val_loss": 0.7110384106636047}
36
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "relu", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.6535947712418301, "f1_macro": 0.5785984020062404, "f1_weighted": 0.5638941158555005, "epochs_completed": 53, "best_epoch": 41, "pred_counts": [0, 42, 16, 95], "classes_predicted": 3, "recall": [0.0, 0.9310344827586207, 0.75, 0.8787878787878788], "final_train_loss": 0.7438108549829114, "best_val_loss": 0.7206449508666992}
37
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "relu", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.7908496732026143, "f1_macro": 0.7933473389355743, "f1_weighted": 0.7922172790684902, "epochs_completed": 80, "best_epoch": 68, "pred_counts": [32, 39, 22, 60], "classes_predicted": 4, "recall": [0.7105263157894737, 0.8620689655172413, 0.9, 0.7727272727272727], "final_train_loss": 0.41584639015950653, "best_val_loss": 0.4509934186935425}
38
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "tanh", "act3": "relu", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.7843137254901961, "f1_macro": 0.7928065044344114, "f1_weighted": 0.7787988606958055, "epochs_completed": 64, "best_epoch": 52, "pred_counts": [27, 34, 23, 69], "classes_predicted": 4, "recall": [0.5789473684210527, 0.896551724137931, 0.95, 0.803030303030303], "final_train_loss": 0.6363272583275511, "best_val_loss": 0.6751461029052734}
39
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "tanh", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.6862745098039216, "f1_macro": 0.6983873894895128, "f1_weighted": 0.6823450785860826, "epochs_completed": 36, "best_epoch": 24, "pred_counts": [31, 40, 19, 63], "classes_predicted": 4, "recall": [0.5, 0.8620689655172413, 0.8, 0.6818181818181818], "final_train_loss": 0.6203715529358178, "best_val_loss": 0.671847403049469}
40
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "tanh", "act3": "tanh", "batch_size": 32, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.7189542483660131, "f1_macro": 0.7369299516908212, "f1_weighted": 0.7149142117394461, "epochs_completed": 64, "best_epoch": 52, "pred_counts": [34, 40, 20, 59], "classes_predicted": 4, "recall": [0.5789473684210527, 0.9310344827586207, 0.85, 0.6666666666666666], "final_train_loss": 0.5900494874569407, "best_val_loss": 0.675346314907074}
41
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "relu", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.7647058823529411, "f1_macro": 0.7727172614946489, "f1_weighted": 0.7689027887730965, "epochs_completed": 56, "best_epoch": 44, "pred_counts": [34, 46, 21, 52], "classes_predicted": 4, "recall": [0.7631578947368421, 0.896551724137931, 0.85, 0.6818181818181818], "final_train_loss": 0.5068238717422151, "best_val_loss": 0.6093516945838928}
42
+ {"key": {"projection_dim": 128, "act1": "tanh", "act2": "relu", "act3": "relu", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.0}, "accuracy": 0.5882352941176471, "f1_macro": 0.540374677002584, "f1_weighted": 0.5054972893550186, "epochs_completed": 31, "best_epoch": 19, "pred_counts": [0, 52, 23, 78], "classes_predicted": 3, "recall": [0.0, 0.9310344827586207, 0.95, 0.6666666666666666], "final_train_loss": 0.716982879136738, "best_val_loss": 0.8123916387557983}
43
+ {"key": {"projection_dim": 128, "act1": "relu", "act2": "tanh", "act3": "tanh", "batch_size": 16, "optimizer": "Adam", "seed": 42}, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.0}, "accuracy": 0.7647058823529411, "f1_macro": 0.7698453537137181, "f1_weighted": 0.7665044761972772, "epochs_completed": 28, "best_epoch": 16, "pred_counts": [42, 39, 21, 51], "classes_predicted": 4, "recall": [0.8157894736842105, 0.8275862068965517, 0.85, 0.6818181818181818], "final_train_loss": 0.5418980226182101, "best_val_loss": 0.6495389342308044}
diagnostics/2026-10-30-other-datasets/results/grid_scTrioseq2_shard0.log ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ [0] 0 {'projection_dim': 64, 'act1': 'relu', 'act2': 'relu', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.722 classes=4
2
+ [0] 2 {'projection_dim': 64, 'act1': 'relu', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.151 classes=1
3
+ [0] 4 {'projection_dim': 64, 'act1': 'relu', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.549 classes=3
4
+ [0] 6 {'projection_dim': 64, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.099 classes=1
5
+ [0] 8 {'projection_dim': 64, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.151 classes=1
6
+ [0] 10 {'projection_dim': 64, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.151 classes=1
7
+ [0] 12 {'projection_dim': 64, 'act1': 'relu', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 16, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.715 classes=4
diagnostics/2026-10-30-other-datasets/results/grid_scTrioseq2_shard1.log ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [1] 1 {'projection_dim': 64, 'act1': 'relu', 'act2': 'relu', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.151 classes=1
2
+ [1] 3 {'projection_dim': 64, 'act1': 'relu', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.151 classes=1
3
+ [1] 5 {'projection_dim': 64, 'act1': 'relu', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.553 classes=4
4
+ [1] 7 {'projection_dim': 64, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.151 classes=1
5
+ [1] 9 {'projection_dim': 64, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.151 classes=1
6
+ [1] 11 {'projection_dim': 64, 'act1': 'relu', 'act2': 'sigmoid', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.151 classes=1
7
+ [1] 13 {'projection_dim': 64, 'act1': 'relu', 'act2': 'tanh', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.151 classes=1
8
+ [1] 15 {'projection_dim': 64, 'act1': 'relu', 'act2': 'tanh', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.151 classes=1
9
+ [1] 17 {'projection_dim': 64, 'act1': 'relu', 'act2': 'tanh', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.749 classes=4
10
+ [1] 19 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.151 classes=1
11
+ [1] 21 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.151 classes=1
12
+ [1] 23 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'relu', 'act3': 'tanh', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.151 classes=1
13
+ [1] 25 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'relu', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.151 classes=1
14
+ [1] 27 {'projection_dim': 64, 'act1': 'sigmoid', 'act2': 'sigmoid', 'act3': 'sigmoid', 'batch_size': 32, 'optimizer': 'Adam', 'seed': 42} -> f1m=0.151 classes=1
diagnostics/2026-10-30-other-datasets/results/sanity_baselines.txt ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ print(f" {name:8s} acc={np.nanmean(r[:,0]):.3f} macroF1={np.nanmean(r[:,1]):.3f} weightedF1={np.nanmean(r[:,2]):.3f} (3 seeds)")
2
+ print(f" {name:8s} acc={np.nanmean(r[:,0]):.3f} macroF1={np.nanmean(r[:,1]):.3f} weightedF1={np.nanmean(r[:,2]):.3f} (3 seeds)")
3
+ print(f" {name:8s} acc={np.nanmean(r[:,0]):.3f} macroF1={np.nanmean(r[:,1]):.3f} weightedF1={np.nanmean(r[:,2]):.3f} (3 seeds)")
4
+ COAD.pkl
5
+ majority acc=0.672 macroF1=nan weightedF1=nan (3 seeds)
6
+ logreg acc=0.505 macroF1=0.219 weightedF1=0.481 (3 seeds)
7
+ forest acc=0.630 macroF1=0.222 weightedF1=0.536 (3 seeds)
8
+ GBM.pkl
9
+ majority acc=0.301 macroF1=nan weightedF1=nan (3 seeds)
10
+ logreg acc=0.781 macroF1=0.808 weightedF1=0.780 (3 seeds)
11
+ forest acc=0.738 macroF1=0.763 weightedF1=0.736 (3 seeds)
12
+ BRCA_singlecell.pkl
13
+ majority acc=0.498 macroF1=nan weightedF1=nan (3 seeds)
14
+ logreg acc=0.746 macroF1=0.690 weightedF1=0.745 (3 seeds)
15
+ forest acc=0.763 macroF1=0.734 weightedF1=0.768 (3 seeds)
diagnostics/2026-10-30-other-datasets/results/seeds_GBM.jsonl ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"config": "D default (lab grid, tanh/relu/tanh, b32)", "seed": 1, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 1, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.47540983606557374, "f1_macro": 0.43307602339181284, "f1_weighted": 0.3943897996357012, "epochs_completed": 42, "pred_counts": [5, 14, 38, 4, 0], "collapsed": false, "recall": [0.13333333333333333, 0.5454545454545454, 0.8947368421052632, 0.8, 0.0]}
2
+ {"config": "D default (lab grid, tanh/relu/tanh, b32)", "seed": 2, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.4262295081967213, "f1_macro": 0.46047619047619043, "f1_weighted": 0.34648711943793914, "epochs_completed": 43, "pred_counts": [6, 21, 0, 5, 29], "collapsed": false, "recall": [0.2, 0.9166666666666666, 0.0, 1.0, 0.6363636363636364]}
3
+ {"config": "D default (lab grid, tanh/relu/tanh, b32)", "seed": 0, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 0, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.5573770491803278, "f1_macro": 0.6098390404162899, "f1_weighted": 0.5592879969261165, "epochs_completed": 62, "pred_counts": [16, 7, 22, 6, 10], "collapsed": false, "recall": [0.4, 0.5, 0.6111111111111112, 1.0, 0.5454545454545454]}
4
+ {"config": "D default (lab grid, tanh/relu/tanh, b32)", "seed": 42, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.5409836065573771, "f1_macro": 0.5952165331475676, "f1_weighted": 0.5327840302402201, "epochs_completed": 44, "pred_counts": [14, 15, 15, 6, 11], "collapsed": false, "recall": [0.4, 0.6666666666666666, 0.3888888888888889, 1.0, 0.6363636363636364]}
5
+ {"config": "D default (lab grid, tanh/relu/tanh, b32)", "seed": 123, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 123, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.6721311475409836, "f1_macro": 0.688671613156739, "f1_weighted": 0.6753076778210889, "epochs_completed": 55, "pred_counts": [23, 11, 15, 4, 8], "collapsed": false, "recall": [0.8, 0.5833333333333334, 0.7222222222222222, 0.8, 0.45454545454545453]}
6
+ {"config": "D default (lab grid, tanh/relu/tanh, b32)", "seed": 3, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 3, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.7704918032786885, "f1_macro": 0.8044513457556937, "f1_weighted": 0.7581712656552285, "epochs_completed": 102, "pred_counts": [8, 12, 24, 5, 12], "collapsed": false, "recall": [0.4, 0.9166666666666666, 0.8333333333333334, 1.0, 0.9090909090909091]}
7
+ {"config": "D default (lab grid, tanh/relu/tanh, b32)", "seed": 7, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 7, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.7049180327868853, "f1_macro": 0.7004285714285714, "f1_weighted": 0.6357962529274006, "epochs_completed": 64, "pred_counts": [1, 10, 31, 5, 14], "collapsed": false, "recall": [0.06666666666666667, 0.8181818181818182, 0.9473684210526315, 1.0, 0.9090909090909091]}
8
+ {"config": "PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03)", "seed": 1, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "SGD", "class_weighting": true, "early_stopping_patience": 12, "seed": 1, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.3114754098360656, "f1_macro": 0.095, "f1_weighted": 0.14795081967213114, "epochs_completed": 28, "pred_counts": [0, 0, 61, 0, 0], "collapsed": true, "recall": [0.0, 0.0, 1.0, 0.0, 0.0]}
9
+ {"config": "PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03)", "seed": 2, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "SGD", "class_weighting": true, "early_stopping_patience": 12, "seed": 2, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 24, "pred_counts": [61, 0, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0, 0.0]}
10
+ {"config": "PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03)", "seed": 0, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "SGD", "class_weighting": true, "early_stopping_patience": 12, "seed": 0, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 18, "pred_counts": [61, 0, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0, 0.0]}
11
+ {"config": "D default (lab grid, tanh/relu/tanh, b32)", "seed": 2024, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2024, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.6557377049180327, "f1_macro": 0.6852649572649573, "f1_weighted": 0.6498080425949277, "epochs_completed": 52, "pred_counts": [11, 14, 18, 4, 14], "collapsed": false, "recall": [0.4, 0.8333333333333334, 0.6666666666666666, 0.8, 0.7272727272727273]}
12
+ {"config": "PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03)", "seed": 42, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "SGD", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 21, "pred_counts": [61, 0, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0, 0.0]}
13
+ {"config": "PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03)", "seed": 7, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "SGD", "class_weighting": true, "early_stopping_patience": 12, "seed": 7, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.3114754098360656, "f1_macro": 0.095, "f1_weighted": 0.14795081967213114, "epochs_completed": 13, "pred_counts": [0, 0, 61, 0, 0], "collapsed": true, "recall": [0.0, 0.0, 1.0, 0.0, 0.0]}
14
+ {"config": "PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03)", "seed": 3, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "SGD", "class_weighting": true, "early_stopping_patience": 12, "seed": 3, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 21, "pred_counts": [0, 0, 61, 0, 0], "collapsed": true, "recall": [0.0, 0.0, 1.0, 0.0, 0.0]}
15
+ {"config": "PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03)", "seed": 123, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "SGD", "class_weighting": true, "early_stopping_patience": 12, "seed": 123, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 37, "pred_counts": [61, 0, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0, 0.0]}
16
+ {"config": "PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03)", "seed": 2024, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "SGD", "class_weighting": true, "early_stopping_patience": 12, "seed": 2024, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 42, "pred_counts": [0, 0, 61, 0, 0], "collapsed": true, "recall": [0.0, 0.0, 1.0, 0.0, 0.0]}
17
+ {"config": "BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep)", "seed": 1, "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 1, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.6065573770491803, "f1_macro": 0.6519371873560633, "f1_weighted": 0.597909189420411, "epochs_completed": 100, "pred_counts": [12, 12, 22, 6, 9], "collapsed": false, "recall": [0.4, 0.9090909090909091, 0.5789473684210527, 1.0, 0.45454545454545453]}
18
+ {"config": "BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep)", "seed": 0, "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 0, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.639344262295082, "f1_macro": 0.6848047538200339, "f1_weighted": 0.6377021347657882, "epochs_completed": 98, "pred_counts": [16, 12, 20, 5, 8], "collapsed": false, "recall": [0.5333333333333333, 0.75, 0.6666666666666666, 1.0, 0.45454545454545453]}
19
+ {"config": "BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep)", "seed": 2, "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.6885245901639344, "f1_macro": 0.7357125632987702, "f1_weighted": 0.6752593553045786, "epochs_completed": 100, "pred_counts": [18, 14, 11, 5, 13], "collapsed": false, "recall": [0.6666666666666666, 1.0, 0.3888888888888889, 1.0, 0.7272727272727273]}
20
+ {"config": "BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep)", "seed": 3, "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 3, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.6229508196721312, "f1_macro": 0.6648750317715836, "f1_weighted": 0.6146956247578069, "epochs_completed": 100, "pred_counts": [10, 17, 19, 6, 9], "collapsed": false, "recall": [0.3333333333333333, 0.75, 0.6111111111111112, 1.0, 0.7272727272727273]}
21
+ {"config": "BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep)", "seed": 7, "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 7, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.5737704918032787, "f1_macro": 0.6060140897293288, "f1_weighted": 0.5766546922202569, "epochs_completed": 97, "pred_counts": [13, 13, 12, 5, 18], "collapsed": false, "recall": [0.4666666666666667, 0.7272727272727273, 0.47368421052631576, 0.8, 0.6363636363636364]}
22
+ {"config": "BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep)", "seed": 42, "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.5573770491803278, "f1_macro": 0.5982382728833033, "f1_weighted": 0.5501787143126551, "epochs_completed": 100, "pred_counts": [19, 14, 11, 6, 11], "collapsed": false, "recall": [0.6, 0.6666666666666666, 0.3888888888888889, 1.0, 0.45454545454545453]}
23
+ {"config": "BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep)", "seed": 123, "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 123, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.639344262295082, "f1_macro": 0.6691968648490387, "f1_weighted": 0.6386531999504202, "epochs_completed": 100, "pred_counts": [12, 11, 19, 6, 13], "collapsed": false, "recall": [0.5333333333333333, 0.6666666666666666, 0.6666666666666666, 1.0, 0.5454545454545454]}
24
+ {"config": "G1 grid b16 p64 tanh/relu/tanh", "seed": 0, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 0, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.639344262295082, "f1_macro": 0.6753141167775314, "f1_weighted": 0.6359577381168745, "epochs_completed": 39, "pred_counts": [13, 9, 23, 6, 10], "collapsed": false, "recall": [0.4666666666666667, 0.5833333333333334, 0.7222222222222222, 1.0, 0.6363636363636364]}
25
+ {"config": "BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep)", "seed": 2024, "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2024, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.5409836065573771, "f1_macro": 0.5938261638411564, "f1_weighted": 0.5299037973848111, "epochs_completed": 92, "pred_counts": [14, 14, 15, 6, 12], "collapsed": false, "recall": [0.3333333333333333, 0.8333333333333334, 0.3888888888888889, 1.0, 0.5454545454545454]}
26
+ {"config": "G1 grid b16 p64 tanh/relu/tanh", "seed": 2, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.5081967213114754, "f1_macro": 0.5089147733825394, "f1_weighted": 0.44799227120361634, "epochs_completed": 41, "pred_counts": [11, 17, 28, 5, 0], "collapsed": false, "recall": [0.26666666666666666, 0.9166666666666666, 0.6111111111111112, 1.0, 0.0]}
27
+ {"config": "G1 grid b16 p64 tanh/relu/tanh", "seed": 1, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 1, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.4426229508196721, "f1_macro": 0.48704381274040714, "f1_weighted": 0.3704265924739964, "epochs_completed": 67, "pred_counts": [4, 11, 0, 6, 40], "collapsed": false, "recall": [0.2, 0.8181818181818182, 0.0, 1.0, 0.9090909090909091]}
28
+ {"config": "G1 grid b16 p64 tanh/relu/tanh", "seed": 7, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 7, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.4426229508196721, "f1_macro": 0.5011815718157181, "f1_weighted": 0.38266826602692255, "epochs_completed": 62, "pred_counts": [12, 14, 0, 5, 30], "collapsed": false, "recall": [0.4, 0.8181818181818182, 0.0, 1.0, 0.6363636363636364]}
29
+ {"config": "G1 grid b16 p64 tanh/relu/tanh", "seed": 42, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.5901639344262295, "f1_macro": 0.6113227081473926, "f1_weighted": 0.5856015283546151, "epochs_completed": 36, "pred_counts": [16, 13, 14, 6, 12], "collapsed": false, "recall": [0.6, 0.6666666666666666, 0.4444444444444444, 0.8, 0.6363636363636364]}
30
+ {"config": "G2 grid b32 p64 tanh/relu/relu", "seed": 0, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 0, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.03}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 23, "pred_counts": [0, 0, 61, 0, 0], "collapsed": true, "recall": [0.0, 0.0, 1.0, 0.0, 0.0]}
31
+ {"config": "G2 grid b32 p64 tanh/relu/relu", "seed": 1, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 1, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.03}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 14, "pred_counts": [61, 0, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0, 0.0]}
32
+ {"config": "G1 grid b16 p64 tanh/relu/tanh", "seed": 123, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 123, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.32786885245901637, "f1_macro": 0.36420168067226893, "f1_weighted": 0.2766221242595399, "epochs_completed": 55, "pred_counts": [13, 20, 0, 5, 23], "collapsed": false, "recall": [0.26666666666666666, 0.6666666666666666, 0.0, 0.8, 0.36363636363636365]}
33
+ {"config": "G1 grid b16 p64 tanh/relu/tanh", "seed": 3, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 3, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.7377049180327869, "f1_macro": 0.7528246820929747, "f1_weighted": 0.717632605077627, "epochs_completed": 70, "pred_counts": [5, 14, 23, 4, 15], "collapsed": false, "recall": [0.3333333333333333, 1.0, 0.7777777777777778, 0.8, 0.9090909090909091]}
34
+ {"config": "G2 grid b32 p64 tanh/relu/relu", "seed": 7, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 7, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.03}, "accuracy": 0.2459016393442623, "f1_macro": 0.07894736842105263, "f1_weighted": 0.09706643658326145, "epochs_completed": 32, "pred_counts": [61, 0, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0, 0.0]}
35
+ {"config": "G2 grid b32 p64 tanh/relu/relu", "seed": 42, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.03}, "accuracy": 0.5573770491803278, "f1_macro": 0.6026491155902921, "f1_weighted": 0.5550033380313033, "epochs_completed": 55, "pred_counts": [15, 14, 16, 5, 11], "collapsed": false, "recall": [0.4, 0.6666666666666666, 0.4444444444444444, 0.8, 0.7272727272727273]}
36
+ {"config": "G2 grid b32 p64 tanh/relu/relu", "seed": 2, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.03}, "accuracy": 0.5737704918032787, "f1_macro": 0.5452063492063491, "f1_weighted": 0.5140775435857402, "epochs_completed": 63, "pred_counts": [10, 16, 31, 4, 0], "collapsed": false, "recall": [0.4, 0.9166666666666666, 0.7777777777777778, 0.8, 0.0]}
37
+ {"config": "G3 grid b16 p128 tanh/tanh/tanh", "seed": 0, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 0, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.6229508196721312, "f1_macro": 0.6716450216450217, "f1_weighted": 0.6048186785891704, "epochs_completed": 44, "pred_counts": [6, 10, 30, 5, 10], "collapsed": false, "recall": [0.2, 0.6666666666666666, 0.7777777777777778, 1.0, 0.7272727272727273]}
38
+ {"config": "G2 grid b32 p64 tanh/relu/relu", "seed": 123, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 123, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.03}, "accuracy": 0.3770491803278688, "f1_macro": 0.42608695652173917, "f1_weighted": 0.368496079828938, "epochs_completed": 65, "pred_counts": [5, 11, 28, 5, 12], "collapsed": false, "recall": [0.13333333333333333, 0.5833333333333334, 0.4444444444444444, 0.8, 0.18181818181818182]}
39
+ {"config": "G1 grid b16 p64 tanh/relu/tanh", "seed": 2024, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2024, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.7213114754098361, "f1_macro": 0.7524021076962254, "f1_weighted": 0.7176338287235107, "epochs_completed": 74, "pred_counts": [11, 9, 16, 6, 19], "collapsed": false, "recall": [0.5333333333333333, 0.75, 0.6111111111111112, 1.0, 1.0]}
40
+ {"config": "G2 grid b32 p64 tanh/relu/relu", "seed": 3, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 3, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.03}, "accuracy": 0.29508196721311475, "f1_macro": 0.09113924050632911, "f1_weighted": 0.13446773189458391, "epochs_completed": 18, "pred_counts": [0, 0, 61, 0, 0], "collapsed": true, "recall": [0.0, 0.0, 1.0, 0.0, 0.0]}
41
+ {"config": "G3 grid b16 p128 tanh/tanh/tanh", "seed": 7, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 7, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.6065573770491803, "f1_macro": 0.6433397129186603, "f1_weighted": 0.5959173268491645, "epochs_completed": 39, "pred_counts": [10, 13, 25, 5, 8], "collapsed": false, "recall": [0.4666666666666667, 0.8181818181818182, 0.6842105263157895, 1.0, 0.2727272727272727]}
42
+ {"config": "G2 grid b32 p64 tanh/relu/relu", "seed": 2024, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2024, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "relu", "sgd_momentum": 0.03}, "accuracy": 0.45901639344262296, "f1_macro": 0.44644403215831785, "f1_weighted": 0.3911767389519146, "epochs_completed": 58, "pred_counts": [9, 15, 31, 6, 0], "collapsed": false, "recall": [0.13333333333333333, 0.75, 0.6666666666666666, 1.0, 0.0]}
43
+ {"config": "G3 grid b16 p128 tanh/tanh/tanh", "seed": 2, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.5573770491803278, "f1_macro": 0.5318747010999523, "f1_weighted": 0.5044374407099905, "epochs_completed": 58, "pred_counts": [19, 12, 23, 7, 0], "collapsed": false, "recall": [0.6666666666666666, 0.75, 0.5555555555555556, 1.0, 0.0]}
44
+ {"config": "G3 grid b16 p128 tanh/tanh/tanh", "seed": 1, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 1, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.7213114754098361, "f1_macro": 0.7603030303030303, "f1_weighted": 0.7188169753743524, "epochs_completed": 98, "pred_counts": [13, 11, 21, 6, 10], "collapsed": false, "recall": [0.5333333333333333, 0.9090909090909091, 0.6842105263157895, 1.0, 0.7272727272727273]}
45
+ {"config": "G3 grid b16 p128 tanh/tanh/tanh", "seed": 42, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.5901639344262295, "f1_macro": 0.5929791912732021, "f1_weighted": 0.5731392264342787, "epochs_completed": 78, "pred_counts": [14, 15, 17, 7, 8], "collapsed": false, "recall": [0.5333333333333333, 0.8333333333333334, 0.6111111111111112, 1.0, 0.18181818181818182]}
46
+ {"config": "G3 grid b16 p128 tanh/tanh/tanh", "seed": 3, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 3, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.7868852459016393, "f1_macro": 0.7962876254180602, "f1_weighted": 0.7849936948297604, "epochs_completed": 92, "pred_counts": [11, 11, 22, 5, 12], "collapsed": false, "recall": [0.6, 0.8333333333333334, 0.8333333333333334, 0.8, 0.9090909090909091]}
47
+ {"config": "G3 grid b16 p128 tanh/tanh/tanh", "seed": 123, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 123, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.7213114754098361, "f1_macro": 0.7444357298474946, "f1_weighted": 0.7206428801028608, "epochs_completed": 84, "pred_counts": [12, 15, 16, 4, 14], "collapsed": false, "recall": [0.6, 0.8333333333333334, 0.6666666666666666, 0.8, 0.8181818181818182]}
48
+ {"config": "G3 grid b16 p128 tanh/tanh/tanh", "seed": 2024, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2024, "biological_activation": "tanh", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.6557377049180327, "f1_macro": 0.704962439521263, "f1_weighted": 0.6595193218867279, "epochs_completed": 55, "pred_counts": [19, 9, 14, 6, 13], "collapsed": false, "recall": [0.6, 0.75, 0.5, 1.0, 0.7272727272727273]}
diagnostics/2026-10-30-other-datasets/results/seeds_GBM_shard0.log ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [0] D default (lab grid, tanh/relu/tanh, b32) seed=0 f1m=0.610
2
+ [0] D default (lab grid, tanh/relu/tanh, b32) seed=7 f1m=0.700
3
+ [0] PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03) seed=0 f1m=0.079
4
+ [0] PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03) seed=7 f1m=0.095
5
+ [0] BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep) seed=0 f1m=0.685
6
+ [0] BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep) seed=7 f1m=0.606
7
+ [0] G1 grid b16 p64 tanh/relu/tanh seed=0 f1m=0.675
8
+ [0] G1 grid b16 p64 tanh/relu/tanh seed=7 f1m=0.501
9
+ [0] G2 grid b32 p64 tanh/relu/relu seed=0 f1m=0.091
10
+ [0] G2 grid b32 p64 tanh/relu/relu seed=7 f1m=0.079
11
+ [0] G3 grid b16 p128 tanh/tanh/tanh seed=0 f1m=0.672
12
+ [0] G3 grid b16 p128 tanh/tanh/tanh seed=7 f1m=0.643
13
+ SHARD 0 DONE
diagnostics/2026-10-30-other-datasets/results/seeds_GBM_shard1.log ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [1] D default (lab grid, tanh/relu/tanh, b32) seed=1 f1m=0.433
2
+ [1] D default (lab grid, tanh/relu/tanh, b32) seed=42 f1m=0.595
3
+ [1] PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03) seed=1 f1m=0.095
4
+ [1] PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03) seed=42 f1m=0.079
5
+ [1] BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep) seed=1 f1m=0.652
6
+ [1] BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep) seed=42 f1m=0.598
7
+ [1] G1 grid b16 p64 tanh/relu/tanh seed=1 f1m=0.487
8
+ [1] G1 grid b16 p64 tanh/relu/tanh seed=42 f1m=0.611
9
+ [1] G2 grid b32 p64 tanh/relu/relu seed=1 f1m=0.079
10
+ [1] G2 grid b32 p64 tanh/relu/relu seed=42 f1m=0.603
11
+ [1] G3 grid b16 p128 tanh/tanh/tanh seed=1 f1m=0.760
12
+ [1] G3 grid b16 p128 tanh/tanh/tanh seed=42 f1m=0.593
13
+ SHARD 1 DONE
diagnostics/2026-10-30-other-datasets/results/seeds_GBM_shard2.log ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [2] D default (lab grid, tanh/relu/tanh, b32) seed=2 f1m=0.460
2
+ [2] D default (lab grid, tanh/relu/tanh, b32) seed=123 f1m=0.689
3
+ [2] PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03) seed=2 f1m=0.079
4
+ [2] PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03) seed=123 f1m=0.079
5
+ [2] BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep) seed=2 f1m=0.736
6
+ [2] BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep) seed=123 f1m=0.669
7
+ [2] G1 grid b16 p64 tanh/relu/tanh seed=2 f1m=0.509
8
+ [2] G1 grid b16 p64 tanh/relu/tanh seed=123 f1m=0.364
9
+ [2] G2 grid b32 p64 tanh/relu/relu seed=2 f1m=0.545
10
+ [2] G2 grid b32 p64 tanh/relu/relu seed=123 f1m=0.426
11
+ [2] G3 grid b16 p128 tanh/tanh/tanh seed=2 f1m=0.532
12
+ [2] G3 grid b16 p128 tanh/tanh/tanh seed=123 f1m=0.744
13
+ SHARD 2 DONE
diagnostics/2026-10-30-other-datasets/results/seeds_GBM_shard3.log ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [3] D default (lab grid, tanh/relu/tanh, b32) seed=3 f1m=0.804
2
+ [3] D default (lab grid, tanh/relu/tanh, b32) seed=2024 f1m=0.685
3
+ [3] PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03) seed=3 f1m=0.091
4
+ [3] PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03) seed=2024 f1m=0.091
5
+ [3] BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep) seed=3 f1m=0.665
6
+ [3] BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep) seed=2024 f1m=0.594
7
+ [3] G1 grid b16 p64 tanh/relu/tanh seed=3 f1m=0.753
8
+ [3] G1 grid b16 p64 tanh/relu/tanh seed=2024 f1m=0.752
9
+ [3] G2 grid b32 p64 tanh/relu/relu seed=3 f1m=0.091
10
+ [3] G2 grid b32 p64 tanh/relu/relu seed=2024 f1m=0.446
11
+ [3] G3 grid b16 p128 tanh/tanh/tanh seed=3 f1m=0.796
12
+ [3] G3 grid b16 p128 tanh/tanh/tanh seed=2024 f1m=0.705
13
+ SHARD 3 DONE
diagnostics/2026-10-30-other-datasets/results/seeds_GBM_summary.txt ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ accuracy macro F1 weighted F1 macro F1 min/max collapsed worst-class recall
2
+ config
3
+ G3 grid b16 p128 tanh/tanh/tanh 0.658 ± 0.079 0.681 ± 0.089 0.645 ± 0.093 0.532 / 0.796 0 0.36
4
+ BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep) 0.609 ± 0.049 0.651 ± 0.049 0.603 ± 0.049 0.594 / 0.736 0 0.41
5
+ D default (lab grid, tanh/relu/tanh, b32) 0.600 ± 0.119 0.622 ± 0.126 0.569 ± 0.141 0.433 / 0.804 0 0.26
6
+ G1 grid b16 p64 tanh/relu/tanh 0.551 ± 0.146 0.582 ± 0.139 0.517 ± 0.170 0.364 / 0.753 0 0.22
7
+ G2 grid b32 p64 tanh/relu/relu 0.381 ± 0.134 0.295 ± 0.231 0.286 ± 0.193 0.079 / 0.603 4 0.07
8
+ PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03) 0.275 ± 0.031 0.086 ± 0.008 0.119 ± 0.024 0.079 / 0.095 8 0.00
diagnostics/2026-10-30-other-datasets/results/seeds_scTrioseq2.jsonl ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"config": "D default (lab grid, tanh/relu/tanh, b32)", "seed": 3, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 3, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.6928104575163399, "f1_macro": 0.7207951274123041, "f1_weighted": 0.6888702127615093, "epochs_completed": 42, "pred_counts": [45, 41, 20, 47], "collapsed": false, "recall": [0.631578947368421, 0.9655172413793104, 0.85, 0.5606060606060606]}
2
+ {"config": "D default (lab grid, tanh/relu/tanh, b32)", "seed": 2, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.8300653594771242, "f1_macro": 0.8439550045793448, "f1_weighted": 0.8307953025557261, "epochs_completed": 48, "pred_counts": [33, 40, 18, 62], "collapsed": false, "recall": [0.7368421052631579, 0.9655172413793104, 0.9, 0.803030303030303]}
3
+ {"config": "D default (lab grid, tanh/relu/tanh, b32)", "seed": 0, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 0, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.8169934640522876, "f1_macro": 0.8298007798103784, "f1_weighted": 0.8170906328057523, "epochs_completed": 66, "pred_counts": [30, 42, 21, 60], "collapsed": false, "recall": [0.7105263157894737, 0.9655172413793104, 0.95, 0.7727272727272727]}
4
+ {"config": "D default (lab grid, tanh/relu/tanh, b32)", "seed": 2024, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2024, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.5686274509803921, "f1_macro": 0.6142857142857143, "f1_weighted": 0.5486150015561779, "epochs_completed": 32, "pred_counts": [46, 43, 25, 39], "collapsed": false, "recall": [0.5, 0.896551724137931, 0.95, 0.3484848484848485]}
5
+ {"config": "D default (lab grid, tanh/relu/tanh, b32)", "seed": 123, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 123, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.7843137254901961, "f1_macro": 0.7930667179662562, "f1_weighted": 0.7778169882484077, "epochs_completed": 32, "pred_counts": [25, 35, 23, 70], "collapsed": false, "recall": [0.5526315789473685, 0.8620689655172413, 1.0, 0.8181818181818182]}
6
+ {"config": "D default (lab grid, tanh/relu/tanh, b32)", "seed": 1, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 1, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.8235294117647058, "f1_macro": 0.8218085785164465, "f1_weighted": 0.8214747925563567, "epochs_completed": 84, "pred_counts": [30, 38, 23, 62], "collapsed": false, "recall": [0.631578947368421, 0.896551724137931, 1.0, 0.8484848484848485]}
7
+ {"config": "PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03)", "seed": 3, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "SGD", "class_weighting": true, "early_stopping_patience": 12, "seed": 3, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 13, "pred_counts": [0, 0, 0, 153], "collapsed": true, "recall": [0.0, 0.0, 0.0, 1.0]}
8
+ {"config": "PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03)", "seed": 2, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "SGD", "class_weighting": true, "early_stopping_patience": 12, "seed": 2, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 23, "pred_counts": [0, 0, 0, 153], "collapsed": true, "recall": [0.0, 0.0, 0.0, 1.0]}
9
+ {"config": "PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03)", "seed": 2024, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "SGD", "class_weighting": true, "early_stopping_patience": 12, "seed": 2024, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 26, "pred_counts": [0, 0, 0, 153], "collapsed": true, "recall": [0.0, 0.0, 0.0, 1.0]}
10
+ {"config": "D default (lab grid, tanh/relu/tanh, b32)", "seed": 7, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 7, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.8562091503267973, "f1_macro": 0.8613644046523303, "f1_weighted": 0.8590313081247338, "epochs_completed": 52, "pred_counts": [38, 39, 17, 59], "collapsed": false, "recall": [0.868421052631579, 0.9310344827586207, 0.85, 0.8181818181818182]}
11
+ {"config": "PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03)", "seed": 123, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "SGD", "class_weighting": true, "early_stopping_patience": 12, "seed": 123, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 27, "pred_counts": [0, 0, 0, 153], "collapsed": true, "recall": [0.0, 0.0, 0.0, 1.0]}
12
+ {"config": "PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03)", "seed": 0, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "SGD", "class_weighting": true, "early_stopping_patience": 12, "seed": 0, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 15, "pred_counts": [0, 0, 0, 153], "collapsed": true, "recall": [0.0, 0.0, 0.0, 1.0]}
13
+ {"config": "PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03)", "seed": 7, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "SGD", "class_weighting": true, "early_stopping_patience": 12, "seed": 7, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 15, "pred_counts": [0, 0, 0, 153], "collapsed": true, "recall": [0.0, 0.0, 0.0, 1.0]}
14
+ {"config": "D default (lab grid, tanh/relu/tanh, b32)", "seed": 42, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.8300653594771242, "f1_macro": 0.8389094286870471, "f1_weighted": 0.8280242914220465, "epochs_completed": 87, "pred_counts": [30, 36, 21, 66], "collapsed": false, "recall": [0.6842105263157895, 0.9310344827586207, 0.95, 0.8333333333333334]}
15
+ {"config": "PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03)", "seed": 1, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "SGD", "class_weighting": true, "early_stopping_patience": 12, "seed": 1, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 22, "pred_counts": [0, 0, 0, 153], "collapsed": true, "recall": [0.0, 0.0, 0.0, 1.0]}
16
+ {"config": "PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03)", "seed": 42, "hyperparameters": {"epochs": 200, "batch_size": 16, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "SGD", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.43137254901960786, "f1_macro": 0.1506849315068493, "f1_weighted": 0.26000537201181845, "epochs_completed": 28, "pred_counts": [0, 0, 0, 153], "collapsed": true, "recall": [0.0, 0.0, 0.0, 1.0]}
17
+ {"config": "BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep)", "seed": 2, "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.8300653594771242, "f1_macro": 0.8342136222910217, "f1_weighted": 0.8319748274954977, "epochs_completed": 67, "pred_counts": [37, 39, 18, 59], "collapsed": false, "recall": [0.7894736842105263, 0.9310344827586207, 0.85, 0.803030303030303]}
18
+ {"config": "BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep)", "seed": 0, "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 0, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.803921568627451, "f1_macro": 0.809327868852459, "f1_weighted": 0.8050590378227794, "epochs_completed": 68, "pred_counts": [42, 32, 20, 59], "collapsed": false, "recall": [0.8421052631578947, 0.7586206896551724, 0.9, 0.7727272727272727]}
19
+ {"config": "BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep)", "seed": 3, "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 3, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.8562091503267973, "f1_macro": 0.85758547008547, "f1_weighted": 0.8586168370482097, "epochs_completed": 100, "pred_counts": [34, 36, 19, 64], "collapsed": false, "recall": [0.8157894736842105, 0.8620689655172413, 0.9, 0.8636363636363636]}
diagnostics/2026-10-30-other-datasets/results/seeds_scTrioseq2_common_shard0.log ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ [0] D default (lab grid, tanh/relu/tanh, b32) seed=0 f1m=0.830
2
+ [0] D default (lab grid, tanh/relu/tanh, b32) seed=7 f1m=0.861
3
+ [0] PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03) seed=0 f1m=0.151
4
+ [0] PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03) seed=7 f1m=0.151
5
+ [0] BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep) seed=0 f1m=0.809
diagnostics/2026-10-30-other-datasets/results/seeds_scTrioseq2_common_shard1.log ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ [1] D default (lab grid, tanh/relu/tanh, b32) seed=1 f1m=0.822
2
+ [1] D default (lab grid, tanh/relu/tanh, b32) seed=42 f1m=0.839
3
+ [1] PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03) seed=1 f1m=0.151
4
+ [1] PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03) seed=42 f1m=0.151
diagnostics/2026-10-30-other-datasets/results/seeds_scTrioseq2_common_shard2.log ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ [2] D default (lab grid, tanh/relu/tanh, b32) seed=2 f1m=0.844
2
+ [2] D default (lab grid, tanh/relu/tanh, b32) seed=123 f1m=0.793
3
+ [2] PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03) seed=2 f1m=0.151
4
+ [2] PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03) seed=123 f1m=0.151
5
+ [2] BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep) seed=2 f1m=0.834
diagnostics/2026-10-30-other-datasets/results/seeds_scTrioseq2_common_shard3.log ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ [3] D default (lab grid, tanh/relu/tanh, b32) seed=3 f1m=0.721
2
+ [3] D default (lab grid, tanh/relu/tanh, b32) seed=2024 f1m=0.614
3
+ [3] PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03) seed=3 f1m=0.151
4
+ [3] PAPER notebook (SGD lr0.01 L2 0.01 b16 relu/sigmoid/tanh, mom 0.03) seed=2024 f1m=0.151
5
+ [3] BASE (Adam lr1e-3 wd1e-4 b16 relu/sigmoid/tanh, 100 ep) seed=3 f1m=0.858
diagnostics/2026-10-30-other-datasets/results/seeds_scTrioseq2_top.jsonl ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ {"config": "S1 grid b32 p128 relu/tanh/tanh", "seed": 1, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 1, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.7189542483660131, "f1_macro": 0.7299061528679224, "f1_weighted": 0.7073511931624367, "epochs_completed": 34, "pred_counts": [23, 45, 20, 65], "collapsed": false, "recall": [0.39473684210526316, 0.9310344827586207, 0.95, 0.7424242424242424]}
2
+ {"config": "S1 grid b32 p128 relu/tanh/tanh", "seed": 0, "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 128, "fusion_dim": 12, "validation_fraction": 0.25, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 0, "biological_activation": "relu", "projection_activation": "tanh", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "accuracy": 0.7254901960784313, "f1_macro": 0.7439041014698909, "f1_weighted": 0.7286905326379011, "epochs_completed": 52, "pred_counts": [36, 51, 18, 48], "collapsed": false, "recall": [0.7368421052631579, 0.9310344827586207, 0.8, 0.6060606060606061]}
diagnostics/2026-10-30-other-datasets/results/seeds_scTrioseq2_top_shard0.log ADDED
@@ -0,0 +1 @@
 
 
1
+ [0] S1 grid b32 p128 relu/tanh/tanh seed=0 f1m=0.744
diagnostics/2026-10-30-other-datasets/results/seeds_scTrioseq2_top_shard1.log ADDED
@@ -0,0 +1 @@
 
 
1
+ [1] S1 grid b32 p128 relu/tanh/tanh seed=1 f1m=0.730
diagnostics/2026-10-30-other-datasets/scripts/build_workspace.py ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Build and cache one example dataset's workspace with the app's own prepare step (auto GenePT).
2
+ Usage: build_workspace.py <dataset> <cache_dir>"""
3
+ import pickle
4
+ import sys
5
+ from pathlib import Path
6
+
7
+ import numpy as np
8
+
9
+ import app
10
+
11
+ DATASET, CACHE = sys.argv[1], Path(sys.argv[2])
12
+ CACHE.mkdir(parents=True, exist_ok=True)
13
+ result = app.prepare_workspace(
14
+ "BioLM-NET examples", DATASET, "", None, None, None, None, None, None, None,
15
+ True, list(app.GENEPT_OPTIONS)[0], progress=lambda *a, **k: None,
16
+ )
17
+ workspace = result[0]
18
+ if workspace is None:
19
+ sys.exit(f"{DATASET}: {result[1]}")
20
+ (CACHE / f"workspace_{DATASET}.pkl").write_bytes(pickle.dumps(workspace))
21
+ counts = np.bincount(workspace.labels)
22
+ print(f"{DATASET}: samples={len(workspace.labels)} classes={workspace.label_names} counts={counts.tolist()} "
23
+ f"majority share={counts.max() / counts.sum():.3f} warnings={workspace.warnings}")
24
+ for name, branch in (("gene", workspace.gene_branch), ("dna", workspace.dna_branch)):
25
+ m = branch.biological_mask
26
+ print(f" {name}: inputs={m.shape[0]} hidden={m.shape[1]} edges={(m > 0).sum()} pathways={len(branch.pathways)}")
diagnostics/2026-10-30-other-datasets/scripts/grid.py ADDED
@@ -0,0 +1,71 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Jubair's hyperparameter grid (email of 2026-10-17) on BRCA, using the Workbench's train() unchanged.
2
+
3
+ Usage: grid.py <workspace.pkl> <out.jsonl> <shard> <n_shards> [seed]
4
+ Resumable: configs already in <out.jsonl> are skipped.
5
+
6
+ Grid as sent, mapped onto Workbench Hyperparameters:
7
+ layer_one -> projection_dim; activation_function_one/two/three -> biological/projection/fusion activation;
8
+ l2_reg -> weight_decay; epoch -> epochs (early stopping, patience 12, still applies);
9
+ SGD gets momentum 0.03 as instructed. num_ex_neuron is not supported by the Workbench (fixed at 0).
10
+ """
11
+ import itertools
12
+ import json
13
+ import sys
14
+ from dataclasses import asdict, replace
15
+ from pathlib import Path
16
+
17
+ import numpy as np
18
+ import pickle
19
+ import torch
20
+ from sklearn.metrics import recall_score
21
+
22
+ import biolmnet.training as training
23
+
24
+ WORKSPACE, OUT, SHARD, SHARDS = Path(sys.argv[1]), Path(sys.argv[2]), int(sys.argv[3]), int(sys.argv[4])
25
+ SEED = int(sys.argv[5]) if len(sys.argv) > 5 else 42
26
+ torch.set_num_threads(int(__import__("os").environ.get("THREADS", "2")))
27
+
28
+ ACT = __import__("os").environ.get("ACTS", "relu,sigmoid,tanh").split(",")
29
+ OPTIMIZERS = ["Adam"] if __import__("os").environ.get("ADAM_ONLY") else ["SGD", "Adam"]
30
+ GRID = list(itertools.product([64, 128], ACT, ACT, ACT, [16, 32], OPTIMIZERS))
31
+
32
+ workspace = pickle.loads(WORKSPACE.read_bytes())
33
+ names = workspace.label_names
34
+ done = set()
35
+ if OUT.exists():
36
+ done = {json.dumps(json.loads(line)["key"]) for line in OUT.read_text().splitlines() if line.strip()}
37
+
38
+ for index, (proj, a1, a2, a3, batch, opt) in enumerate(GRID):
39
+ if index % SHARDS != SHARD:
40
+ continue
41
+ key = {"projection_dim": proj, "act1": a1, "act2": a2, "act3": a3, "batch_size": batch, "optimizer": opt, "seed": SEED}
42
+ if json.dumps(key) in done:
43
+ continue
44
+ hp = replace(
45
+ training.Hyperparameters(),
46
+ epochs=200, batch_size=batch, learning_rate=0.01, weight_decay=0.01, dropout=0.3,
47
+ projection_dim=proj, optimizer=opt, sgd_momentum=0.03 if opt == "SGD" else 0.0,
48
+ biological_activation=a1, projection_activation=a2, fusion_activation=a3, seed=SEED,
49
+ )
50
+ try:
51
+ result = training.train(workspace, hp)
52
+ except Exception as exc: # record and keep going
53
+ row = {"key": key, "error": repr(exc)}
54
+ else:
55
+ m = result.bundle.metrics
56
+ vp = result.validation_predictions
57
+ counts = vp["predicted"].value_counts().reindex(names, fill_value=0).astype(int).tolist()
58
+ history = result.bundle.history
59
+ best_epoch = int(min(history, key=lambda h: h["validation_loss"])["epoch"])
60
+ row = {
61
+ "key": key, "hyperparameters": asdict(hp),
62
+ "accuracy": m["accuracy"], "f1_macro": m["f1_macro"], "f1_weighted": m["f1_weighted"],
63
+ "epochs_completed": m["epochs_completed"], "best_epoch": best_epoch,
64
+ "pred_counts": counts, "classes_predicted": int((np.array(counts) > 0).sum()),
65
+ "recall": recall_score(vp["observed"], vp["predicted"], labels=names, average=None, zero_division=0).tolist(),
66
+ "final_train_loss": history[-1]["training_loss"], "best_val_loss": m["validation_loss"],
67
+ }
68
+ with OUT.open("a") as handle:
69
+ handle.write(json.dumps(row) + "\n")
70
+ print(f"[{SHARD}] {index} {key} -> " + (row.get("error") or f"f1m={row['f1_macro']:.3f} classes={row['classes_predicted']}"), flush=True)
71
+ print(f"SHARD {SHARD} DONE", flush=True)
diagnostics/2026-10-30-other-datasets/scripts/grid_summary.py ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Summarise grid.py output. Usage: grid_summary.py <grid.jsonl>"""
2
+ import json
3
+ import sys
4
+
5
+ import pandas as pd
6
+
7
+ rows = []
8
+ for line in open(sys.argv[1]):
9
+ r = json.loads(line)
10
+ if "error" in r:
11
+ print("ERROR", r["key"], r["error"])
12
+ continue
13
+ rows.append({**r["key"], "acc": r["accuracy"], "f1_macro": r["f1_macro"], "f1_w": r["f1_weighted"],
14
+ "epochs": r["epochs_completed"], "best": r["best_epoch"], "classes": r["classes_predicted"],
15
+ "collapsed": r["classes_predicted"] == 1, "pred": "/".join(map(str, r["pred_counts"])),
16
+ "min_recall": min(r["recall"])})
17
+ df = pd.DataFrame(rows)
18
+ df["acts"] = df.act1 + "/" + df.act2 + "/" + df.act3
19
+ pd.set_option("display.width", 220)
20
+ pd.set_option("display.max_rows", 300)
21
+
22
+ print(f"configs: {len(df)} collapsed (1 class): {df.collapsed.sum()} "
23
+ f"all 5 classes predicted: {(df.classes == 5).sum()} macro F1 >= 0.70: {(df.f1_macro >= 0.70).sum()}")
24
+ print("\nTOP 20 by macro F1")
25
+ print(df.sort_values("f1_macro", ascending=False).head(20)[
26
+ ["optimizer", "batch_size", "projection_dim", "acts", "acc", "f1_macro", "f1_w", "min_recall", "best", "epochs", "pred"]
27
+ ].round(3).to_string(index=False))
28
+ for factor in ["optimizer", "batch_size", "projection_dim", "act1", "act2", "act3"]:
29
+ g = df.groupby(factor).agg(n=("f1_macro", "size"), f1_mean=("f1_macro", "mean"), f1_max=("f1_macro", "max"),
30
+ collapsed=("collapsed", "sum"), working=("f1_macro", lambda s: int((s >= 0.70).sum())))
31
+ print(f"\nBY {factor}\n" + g.round(3).to_string())
diagnostics/2026-10-30-other-datasets/scripts/sanity_baselines.py ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Label-alignment sanity check: classical models on the Workbench's own split and scaling.
2
+ If simple models beat the majority-class rate clearly, labels carry signal for these rows.
3
+ Usage: sanity_baselines.py <workspace.pkl> [<workspace.pkl> ...]"""
4
+ import pickle
5
+ import sys
6
+
7
+ import numpy as np
8
+ from sklearn.ensemble import RandomForestClassifier
9
+ from sklearn.linear_model import LogisticRegression
10
+ from sklearn.metrics import accuracy_score, f1_score
11
+
12
+ from biolmnet.training import split_train_validation
13
+
14
+ for path in sys.argv[1:]:
15
+ ws = pickle.load(open(path, "rb"))
16
+ X = np.hstack([ws.gene_expression, ws.dna_methylation]).astype(np.float64)
17
+ y = ws.labels
18
+ rows = []
19
+ for seed in (0, 1, 42):
20
+ tr, va = split_train_validation(y, 0.25, seed)
21
+ mean, std = X[tr].mean(0), X[tr].std(0) + 1e-8
22
+ Xtr, Xva = (X[tr] - mean) / std, (X[va] - mean) / std
23
+ majority = np.bincount(y[tr]).argmax()
24
+ for name, model in (("logreg", LogisticRegression(max_iter=3000, class_weight="balanced", C=0.1)),
25
+ ("forest", RandomForestClassifier(500, class_weight="balanced", random_state=seed, n_jobs=2))):
26
+ p = model.fit(Xtr, y[tr]).predict(Xva)
27
+ rows.append((name, accuracy_score(y[va], p), f1_score(y[va], p, average="macro"), f1_score(y[va], p, average="weighted")))
28
+ rows.append(("majority", accuracy_score(y[va], np.full(len(va), majority)), np.nan, np.nan))
29
+ print(path.split("workspace_")[-1])
30
+ for name in ("majority", "logreg", "forest"):
31
+ r = np.array([x[1:] for x in rows if x[0] == name], dtype=float)
32
+ print(f" {name:8s} acc={np.nanmean(r[:,0]):.3f} macroF1={np.nanmean(r[:,1]):.3f} weightedF1={np.nanmean(r[:,2]):.3f} (3 seeds)")
diagnostics/2026-10-30-other-datasets/scripts/seed_check.py ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Run named configurations across 8 seeds on one workspace.
2
+ Usage: seed_check.py <workspace.pkl> <configs.json> <out.jsonl> <shard> <n_shards>
3
+ configs.json maps a name to Hyperparameters overrides. Resumable: (config, seed) pairs already in out.jsonl are skipped."""
4
+ import json
5
+ import os
6
+ import pickle
7
+ import sys
8
+ from dataclasses import asdict, replace
9
+ from pathlib import Path
10
+
11
+ import numpy as np
12
+ import torch
13
+ from sklearn.metrics import recall_score
14
+
15
+ import biolmnet.training as training
16
+
17
+ WORKSPACE, CONFIGS, OUT = Path(sys.argv[1]), json.loads(Path(sys.argv[2]).read_text()), Path(sys.argv[3])
18
+ SHARD, SHARDS = int(sys.argv[4]), int(sys.argv[5])
19
+ torch.set_num_threads(int(os.environ.get("THREADS", "2")))
20
+ SEEDS = [0, 1, 2, 3, 7, 42, 123, 2024]
21
+
22
+ workspace = pickle.loads(WORKSPACE.read_bytes())
23
+ names = workspace.label_names
24
+ done = set()
25
+ if OUT.exists():
26
+ done = {(r["config"], r["seed"]) for r in map(json.loads, OUT.read_text().splitlines()) if r}
27
+
28
+ jobs = [(c, s) for c in CONFIGS for s in SEEDS]
29
+ for index, (config, seed) in enumerate(jobs):
30
+ if index % SHARDS != SHARD or (config, seed) in done:
31
+ continue
32
+ hp = replace(training.Hyperparameters(), seed=seed, **CONFIGS[config])
33
+ try:
34
+ result = training.train(workspace, hp)
35
+ except Exception as exc:
36
+ row = {"config": config, "seed": seed, "error": repr(exc)}
37
+ else:
38
+ m = result.bundle.metrics
39
+ vp = result.validation_predictions
40
+ counts = vp["predicted"].value_counts().reindex(names, fill_value=0).astype(int).tolist()
41
+ row = {
42
+ "config": config, "seed": seed, "hyperparameters": asdict(hp),
43
+ "accuracy": m["accuracy"], "f1_macro": m["f1_macro"], "f1_weighted": m["f1_weighted"],
44
+ "epochs_completed": m["epochs_completed"],
45
+ "pred_counts": counts, "collapsed": int((np.array(counts) > 0).sum()) == 1,
46
+ "recall": recall_score(vp["observed"], vp["predicted"], labels=names, average=None, zero_division=0).tolist(),
47
+ }
48
+ with OUT.open("a") as handle:
49
+ handle.write(json.dumps(row) + "\n")
50
+ print(f"[{SHARD}] {config} seed={seed} " + (row.get("error") or f"f1m={row['f1_macro']:.3f}"), flush=True)
51
+ print(f"SHARD {SHARD} DONE", flush=True)
diagnostics/2026-10-30-other-datasets/scripts/seed_summary.py ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Summarise seed_check.py output. Usage: seed_summary.py <seeds.jsonl>"""
2
+ import json
3
+ import sys
4
+
5
+ import pandas as pd
6
+
7
+ rows = [json.loads(l) for l in open(sys.argv[1])]
8
+ errors = [r for r in rows if "error" in r]
9
+ df = pd.DataFrame([r for r in rows if "error" not in r])
10
+ df["min_recall"] = df.recall.map(min)
11
+ g = df.groupby("config")
12
+ s = pd.DataFrame({
13
+ "accuracy": g.accuracy.mean().map("{:.3f}".format) + " ± " + g.accuracy.std().map("{:.3f}".format),
14
+ "macro F1": g.f1_macro.mean().map("{:.3f}".format) + " ± " + g.f1_macro.std().map("{:.3f}".format),
15
+ "weighted F1": g.f1_weighted.mean().map("{:.3f}".format) + " ± " + g.f1_weighted.std().map("{:.3f}".format),
16
+ "macro F1 min/max": g.f1_macro.min().map("{:.3f}".format) + " / " + g.f1_macro.max().map("{:.3f}".format),
17
+ "collapsed": g.collapsed.sum(), "worst-class recall": g.min_recall.mean().round(2),
18
+ "_sort": g.f1_macro.mean(),
19
+ }).sort_values("_sort", ascending=False).drop(columns="_sort")
20
+ pd.set_option("display.width", 250); pd.set_option("display.max_colwidth", 70)
21
+ print(s.to_string())
22
+ for r in errors:
23
+ print("ERROR", r["config"], r["seed"], r["error"])
diagnostics/2026-11-06-test-split-reproduction/README.md ADDED
@@ -0,0 +1,36 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Paper reproduction on the 80/20 test split (2026-11-06)
2
+
3
+ All scores are on the untouched 20% test set, averaged over 10 seeds (the paper reports 10 runs). Early stopping uses 1/8 of the training part. The variants marked "notebook first layer" swap in, for these runs only, the upstream notebook's first layer: ReLU(xW + b) followed by a fixed multiplication by the biological mask. In the Workbench, the mask is applied to the trainable weights (W ⊙ M) instead. `biolmnet/model.py` was not changed.
4
+
5
+ Configurations:
6
+
7
+ - **D:** the current Workbench default, from the lab grid. Adam, lr 0.01, L2 0.01, batch 32, tanh/relu/tanh activations.
8
+ - **BASE:** Adam, lr 0.001, weight decay 1e-4, batch 16, published relu/sigmoid/tanh activations, 100 epochs.
9
+ - **Notebook replica:** the notebook's first layer with the paper notebook's SGD settings (lr 0.01, L2 0.01, batch 16, relu/sigmoid/tanh, 200 epochs, momentum 0.03).
10
+
11
+ | Dataset | Paper | D | BASE | BASE + notebook first layer |
12
+ |---|---|---|---|---|
13
+ | scTrioseq2 | acc 0.95, F1 0.95 | macro F1 0.756 ± 0.031 | 0.826 ± 0.027 | **acc 0.935 ± 0.021, macro F1 0.937 ± 0.020** |
14
+ | BRCA | F1 0.82 | 0.720 ± 0.036 | 0.726 ± 0.032 | **macro F1 0.762 ± 0.027 (weighted 0.796)** |
15
+ | GBM | not stated in text | 0.481 ± 0.138 | 0.608 ± 0.081 | **0.785 ± 0.067** |
16
+ | COAD | acc 0.87 | about 0.21 | about 0.20 | about 0.21 |
17
+
18
+ Other runs:
19
+
20
+ - On scTrioseq2, the notebook replica reaches macro F1 0.910 ± 0.025. The same SGD settings collapse with the Workbench's first layer.
21
+ - D with the notebook first layer scores 0.359 on scTrioseq2, so the lab-grid lr 0.01 and L2 0.01 don't suit that layer.
22
+
23
+ ## Conclusions
24
+
25
+ - The first-layer formulation, not the hyperparameters, explains most of the gap to the paper. The BASE settings with the notebook first layer come within about 0.015 of the paper on scTrioseq2.
26
+ - The current default D is the weakest option on every dataset.
27
+ - COAD can't be evaluated. Its `label.csv` has 260 rows against 256 omics rows, and simple classifiers score below the majority-class rate, so the labels don't match the samples (see `diagnostics/2026-10-30-other-datasets`).
28
+
29
+ ## Files
30
+
31
+ - `scripts/run_seeds.py <workspace.pkl> <configs.json> <out.jsonl> <shard> <n_shards> [seeds]`: runs the seed experiments.
32
+ - `scripts/summary.py <results.jsonl> ...`: prints the summary tables.
33
+ - `results/configs_*.json`: the configurations used.
34
+ - `results/<dataset>.jsonl`: per-run results.
35
+ - `results/<dataset>_summary.txt`: summary tables.
36
+ - Workspaces are built with `diagnostics/2026-10-30-other-datasets/scripts/build_workspace.py <dataset> <cache_dir>`.
diagnostics/2026-11-06-test-split-reproduction/results/BRCA.jsonl ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"config": "D default", "seed": 0, "first_layer": "masked_weights", "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 0, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.7657142857142857, "f1_macro": 0.7439761687651749, "f1_weighted": 0.7744612874011866, "precision_macro": 0.7690821256038647, "recall_macro": 0.7526857084278374, "epochs_completed": 32, "pred_counts": [38, 23, 6, 72, 36], "collapsed": false, "recall": [0.8260869565217391, 0.8461538461538461, 0.5555555555555556, 0.735632183908046, 0.8]}
2
+ {"config": "D default", "seed": 1, "first_layer": "masked_weights", "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 1, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.7657142857142857, "f1_macro": 0.6759025837512732, "f1_weighted": 0.7657823862954246, "precision_macro": 0.7876636034530773, "recall_macro": 0.6892635733415344, "epochs_completed": 29, "pred_counts": [42, 27, 2, 76, 28], "collapsed": false, "recall": [0.9130434782608695, 0.8846153846153846, 0.2222222222222222, 0.7931034482758621, 0.6333333333333333]}
3
+ {"config": "D default", "seed": 2, "first_layer": "masked_weights", "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.8114285714285714, "f1_macro": 0.7931423407972349, "f1_weighted": 0.8242114013510994, "precision_macro": 0.7855668855668856, "recall_macro": 0.8410087264060279, "epochs_completed": 48, "pred_counts": [39, 21, 14, 74, 27], "collapsed": false, "recall": [0.8260869565217391, 0.8076923076923077, 1.0, 0.8045977011494253, 0.7666666666666667]}
4
+ {"config": "D default", "seed": 3, "first_layer": "masked_weights", "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 3, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.7028571428571428, "f1_macro": 0.6854015689284652, "f1_weighted": 0.7129273789673773, "precision_macro": 0.7492819339318195, "recall_macro": 0.6869713861018208, "epochs_completed": 57, "pred_counts": [38, 25, 4, 69, 39], "collapsed": false, "recall": [0.7391304347826086, 0.8846153846153846, 0.4444444444444444, 0.6666666666666666, 0.7]}
5
+ {"config": "D default", "seed": 7, "first_layer": "masked_weights", "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 7, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.7485714285714286, "f1_macro": 0.7070707070707071, "f1_weighted": 0.7620531849103278, "precision_macro": 0.7081360974031091, "recall_macro": 0.7380568690013967, "epochs_completed": 56, "pred_counts": [43, 23, 9, 67, 33], "collapsed": false, "recall": [0.8695652173913043, 0.8076923076923077, 0.5555555555555556, 0.7241379310344828, 0.7333333333333333]}
6
+ {"config": "D default", "seed": 42, "first_layer": "masked_weights", "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.7657142857142857, "f1_macro": 0.7492861748180897, "f1_weighted": 0.774858721825287, "precision_macro": 0.7457303457303457, "recall_macro": 0.7788264841938005, "epochs_completed": 38, "pred_counts": [37, 21, 12, 78, 27], "collapsed": false, "recall": [0.782608695652174, 0.8076923076923077, 0.8888888888888888, 0.7816091954022989, 0.6333333333333333]}
7
+ {"config": "D default", "seed": 123, "first_layer": "masked_weights", "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 123, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.7542857142857143, "f1_macro": 0.7201195060134211, "f1_weighted": 0.7634685266063347, "precision_macro": 0.7472849614153961, "recall_macro": 0.7293035533515292, "epochs_completed": 33, "pred_counts": [44, 24, 7, 77, 23], "collapsed": false, "recall": [0.8695652173913043, 0.8846153846153846, 0.5555555555555556, 0.7701149425287356, 0.5666666666666667]}
8
+ {"config": "D default", "seed": 2024, "first_layer": "masked_weights", "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2024, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.7485714285714286, "f1_macro": 0.7016661557387966, "f1_weighted": 0.7539214248225028, "precision_macro": 0.7418721069329589, "recall_macro": 0.6974505055164726, "epochs_completed": 28, "pred_counts": [39, 24, 8, 87, 17], "collapsed": false, "recall": [0.6956521739130435, 0.8076923076923077, 0.6666666666666666, 0.8505747126436781, 0.4666666666666667]}
9
+ {"config": "D default", "seed": 5, "first_layer": "masked_weights", "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 5, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.6971428571428572, "f1_macro": 0.689902352745196, "f1_weighted": 0.7148133916235159, "precision_macro": 0.6850793650793651, "recall_macro": 0.7404756596060944, "epochs_completed": 45, "pred_counts": [45, 22, 14, 66, 28], "collapsed": false, "recall": [0.7391304347826086, 0.8076923076923077, 0.8888888888888888, 0.6666666666666666, 0.6]}
10
+ {"config": "D default", "seed": 11, "first_layer": "masked_weights", "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 11, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.76, "f1_macro": 0.7352592131058812, "f1_weighted": 0.7704836446663403, "precision_macro": 0.7287831537831538, "recall_macro": 0.7692845884749933, "epochs_completed": 33, "pred_counts": [39, 24, 9, 66, 37], "collapsed": false, "recall": [0.8260869565217391, 0.8076923076923077, 0.6666666666666666, 0.7126436781609196, 0.8333333333333334]}
11
+ {"config": "BASE lr1e-3 wd1e-4", "seed": 0, "first_layer": "masked_weights", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 0, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.7657142857142857, "f1_macro": 0.723169147176943, "f1_weighted": 0.7732641490331073, "precision_macro": 0.7304852320675106, "recall_macro": 0.7281018465126412, "epochs_completed": 83, "pred_counts": [33, 22, 8, 79, 33], "collapsed": false, "recall": [0.6956521739130435, 0.8076923076923077, 0.5555555555555556, 0.7816091954022989, 0.8]}
12
+ {"config": "BASE lr1e-3 wd1e-4", "seed": 1, "first_layer": "masked_weights", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 1, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.7657142857142857, "f1_macro": 0.7295344971397809, "f1_weighted": 0.7765742951286274, "precision_macro": 0.7356186486621269, "recall_macro": 0.7433475569907354, "epochs_completed": 59, "pred_counts": [39, 23, 9, 77, 27], "collapsed": false, "recall": [0.782608695652174, 0.8076923076923077, 0.6666666666666666, 0.7931034482758621, 0.6666666666666666]}
13
+ {"config": "BASE lr1e-3 wd1e-4", "seed": 2, "first_layer": "masked_weights", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.7885714285714286, "f1_macro": 0.7661212712908659, "f1_weighted": 0.7994210189745566, "precision_macro": 0.75359754896597, "recall_macro": 0.8085700739373903, "epochs_completed": 84, "pred_counts": [38, 25, 13, 74, 25], "collapsed": false, "recall": [0.782608695652174, 0.9230769230769231, 0.8888888888888888, 0.7816091954022989, 0.6666666666666666]}
14
+ {"config": "BASE lr1e-3 wd1e-4", "seed": 3, "first_layer": "masked_weights", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 3, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.6914285714285714, "f1_macro": 0.6681744271988175, "f1_weighted": 0.6976819278282693, "precision_macro": 0.7298090798090798, "recall_macro": 0.6585056189853791, "epochs_completed": 82, "pred_counts": [32, 26, 4, 77, 36], "collapsed": false, "recall": [0.6956521739130435, 0.8846153846153846, 0.4444444444444444, 0.7011494252873564, 0.5666666666666667]}
15
+ {"config": "BASE lr1e-3 wd1e-4", "seed": 7, "first_layer": "masked_weights", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 7, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.76, "f1_macro": 0.7440209632239994, "f1_weighted": 0.7735428623062127, "precision_macro": 0.7939330065359478, "recall_macro": 0.752415843360371, "epochs_completed": 100, "pred_counts": [45, 25, 5, 68, 32], "collapsed": false, "recall": [0.8695652173913043, 0.8461538461538461, 0.5555555555555556, 0.7241379310344828, 0.7666666666666667]}
16
+ {"config": "BASE lr1e-3 wd1e-4", "seed": 42, "first_layer": "masked_weights", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.7714285714285715, "f1_macro": 0.7511083718541346, "f1_weighted": 0.7789654351595475, "precision_macro": 0.7439566755083996, "recall_macro": 0.7699598918489473, "epochs_completed": 67, "pred_counts": [34, 24, 10, 78, 29], "collapsed": false, "recall": [0.7391304347826086, 0.8846153846153846, 0.7777777777777778, 0.7816091954022989, 0.6666666666666666]}
17
+ {"config": "BASE lr1e-3 wd1e-4", "seed": 123, "first_layer": "masked_weights", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 123, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.7428571428571429, "f1_macro": 0.736214405095514, "f1_weighted": 0.7546953012010187, "precision_macro": 0.7446586879432624, "recall_macro": 0.7705262753238765, "epochs_completed": 63, "pred_counts": [47, 24, 8, 64, 32], "collapsed": false, "recall": [0.9565217391304348, 0.8846153846153846, 0.6666666666666666, 0.6781609195402298, 0.6666666666666666]}
18
+ {"config": "BASE lr1e-3 wd1e-4", "seed": 2024, "first_layer": "masked_weights", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2024, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.7028571428571428, "f1_macro": 0.6746308798047022, "f1_weighted": 0.7178134086700942, "precision_macro": 0.6763854425144747, "recall_macro": 0.690134163687387, "epochs_completed": 43, "pred_counts": [39, 24, 9, 72, 31], "collapsed": false, "recall": [0.6521739130434783, 0.8076923076923077, 0.6666666666666666, 0.7241379310344828, 0.6]}
19
+ {"config": "BASE lr1e-3 wd1e-4", "seed": 5, "first_layer": "masked_weights", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 5, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.7371428571428571, "f1_macro": 0.7239937823440212, "f1_weighted": 0.7518301531247148, "precision_macro": 0.7429420289855072, "recall_macro": 0.7392757467420136, "epochs_completed": 66, "pred_counts": [46, 24, 8, 72, 25], "collapsed": false, "recall": [0.8260869565217391, 0.8461538461538461, 0.6666666666666666, 0.7241379310344828, 0.6333333333333333]}
20
+ {"config": "BASE lr1e-3 wd1e-4", "seed": 11, "first_layer": "masked_weights", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 11, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.7542857142857143, "f1_macro": 0.7455712704303955, "f1_weighted": 0.7641840784622984, "precision_macro": 0.7350595238095238, "recall_macro": 0.7903645613090889, "epochs_completed": 50, "pred_counts": [40, 21, 12, 70, 32], "collapsed": false, "recall": [0.8695652173913043, 0.7692307692307693, 0.8888888888888888, 0.7241379310344828, 0.7]}
21
+ {"config": "BASE + notebook first layer", "seed": 0, "first_layer": "dense_then_mask", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 0, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.8057142857142857, "f1_macro": 0.7486689105350364, "f1_weighted": 0.8098612601539289, "precision_macro": 0.7727109497199451, "recall_macro": 0.7459129409654147, "epochs_completed": 27, "pred_counts": [35, 23, 8, 87, 22], "collapsed": false, "recall": [0.8260869565217391, 0.8076923076923077, 0.5555555555555556, 0.8735632183908046, 0.6666666666666666]}
22
+ {"config": "BASE + notebook first layer", "seed": 1, "first_layer": "dense_then_mask", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 1, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.8057142857142857, "f1_macro": 0.804470792148277, "f1_weighted": 0.8159936388348442, "precision_macro": 0.8251428571428571, "recall_macro": 0.8116790322787324, "epochs_completed": 20, "pred_counts": [40, 25, 7, 75, 28], "collapsed": false, "recall": [0.8695652173913043, 0.8846153846153846, 0.7777777777777778, 0.7931034482758621, 0.7333333333333333]}
23
+ {"config": "BASE + notebook first layer", "seed": 2, "first_layer": "dense_then_mask", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.8114285714285714, "f1_macro": 0.802543539409899, "f1_weighted": 0.8229224626591317, "precision_macro": 0.7916146916146917, "recall_macro": 0.8474278245492638, "epochs_completed": 26, "pred_counts": [39, 23, 13, 74, 26], "collapsed": false, "recall": [0.8260869565217391, 0.8846153846153846, 1.0, 0.7931034482758621, 0.7333333333333333]}
24
+ {"config": "BASE + notebook first layer", "seed": 3, "first_layer": "dense_then_mask", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 3, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.7657142857142857, "f1_macro": 0.7449397815671768, "f1_weighted": 0.7756279760656937, "precision_macro": 0.7904485538169749, "recall_macro": 0.7429467317623241, "epochs_completed": 29, "pred_counts": [38, 25, 5, 74, 33], "collapsed": false, "recall": [0.782608695652174, 0.8846153846153846, 0.5555555555555556, 0.7586206896551724, 0.7333333333333333]}
25
+ {"config": "BASE + notebook first layer", "seed": 7, "first_layer": "dense_then_mask", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 7, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.7942857142857143, "f1_macro": 0.7324595729473778, "f1_weighted": 0.8045701012112162, "precision_macro": 0.7693969093969095, "recall_macro": 0.7294340009482438, "epochs_completed": 27, "pred_counts": [37, 26, 5, 77, 30], "collapsed": false, "recall": [0.6956521739130435, 0.8461538461538461, 0.4444444444444444, 0.8275862068965517, 0.8333333333333334]}
26
+ {"config": "BASE + notebook first layer", "seed": 42, "first_layer": "dense_then_mask", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.7942857142857143, "f1_macro": 0.772506264805419, "f1_weighted": 0.8051051192818919, "precision_macro": 0.7638562091503268, "recall_macro": 0.8006845295301067, "epochs_completed": 20, "pred_counts": [36, 23, 10, 72, 34], "collapsed": false, "recall": [0.8260869565217391, 0.8846153846153846, 0.7777777777777778, 0.7816091954022989, 0.7333333333333333]}
27
+ {"config": "BASE + notebook first layer", "seed": 123, "first_layer": "dense_then_mask", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 123, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.76, "f1_macro": 0.7653114989585579, "f1_weighted": 0.7733159971680982, "precision_macro": 0.772829131652661, "recall_macro": 0.7868914260818309, "epochs_completed": 24, "pred_counts": [42, 24, 8, 68, 33], "collapsed": false, "recall": [0.8260869565217391, 0.8846153846153846, 0.7777777777777778, 0.7126436781609196, 0.7333333333333333]}
28
+ {"config": "BASE + notebook first layer", "seed": 2024, "first_layer": "dense_then_mask", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2024, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.7885714285714286, "f1_macro": 0.7301927208914126, "f1_weighted": 0.798480521444571, "precision_macro": 0.743921448921449, "recall_macro": 0.7438073271056779, "epochs_completed": 25, "pred_counts": [39, 24, 10, 81, 21], "collapsed": false, "recall": [0.782608695652174, 0.8076923076923077, 0.6666666666666666, 0.8620689655172413, 0.6]}
29
+ {"config": "BASE + notebook first layer", "seed": 5, "first_layer": "dense_then_mask", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 5, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.7428571428571429, "f1_macro": 0.7404041515384897, "f1_weighted": 0.7535813867851026, "precision_macro": 0.7227756452843188, "recall_macro": 0.7958379784466741, "epochs_completed": 20, "pred_counts": [38, 23, 12, 62, 40], "collapsed": false, "recall": [0.782608695652174, 0.8076923076923077, 0.8888888888888888, 0.6666666666666666, 0.8333333333333334]}
30
+ {"config": "BASE + notebook first layer", "seed": 11, "first_layer": "dense_then_mask", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 11, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 175, "accuracy": 0.7942857142857143, "f1_macro": 0.7783553678827129, "f1_weighted": 0.8033745148460598, "precision_macro": 0.7791058201058201, "recall_macro": 0.791233101398019, "epochs_completed": 37, "pred_counts": [35, 27, 8, 75, 30], "collapsed": false, "recall": [0.7391304347826086, 0.8461538461538461, 0.7777777777777778, 0.7931034482758621, 0.8]}
diagnostics/2026-11-06-test-split-reproduction/results/BRCA_shard0.log ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [0] D default seed=0 acc=0.766 f1w=0.774
2
+ [0] D default seed=1 acc=0.766 f1w=0.766
3
+ [0] D default seed=2 acc=0.811 f1w=0.824
4
+ [0] D default seed=3 acc=0.703 f1w=0.713
5
+ [0] D default seed=7 acc=0.749 f1w=0.762
6
+ [0] D default seed=42 acc=0.766 f1w=0.775
7
+ [0] D default seed=123 acc=0.754 f1w=0.763
8
+ [0] D default seed=2024 acc=0.749 f1w=0.754
9
+ [0] D default seed=5 acc=0.697 f1w=0.715
10
+ [0] D default seed=11 acc=0.760 f1w=0.770
11
+ [0] BASE lr1e-3 wd1e-4 seed=0 acc=0.766 f1w=0.773
12
+ [0] BASE lr1e-3 wd1e-4 seed=1 acc=0.766 f1w=0.777
13
+ [0] BASE lr1e-3 wd1e-4 seed=2 acc=0.789 f1w=0.799
14
+ [0] BASE lr1e-3 wd1e-4 seed=3 acc=0.691 f1w=0.698
15
+ [0] BASE lr1e-3 wd1e-4 seed=7 acc=0.760 f1w=0.774
16
+ [0] BASE lr1e-3 wd1e-4 seed=42 acc=0.771 f1w=0.779
17
+ [0] BASE lr1e-3 wd1e-4 seed=123 acc=0.743 f1w=0.755
18
+ [0] BASE lr1e-3 wd1e-4 seed=2024 acc=0.703 f1w=0.718
19
+ [0] BASE lr1e-3 wd1e-4 seed=5 acc=0.737 f1w=0.752
20
+ [0] BASE lr1e-3 wd1e-4 seed=11 acc=0.754 f1w=0.764
21
+ [0] BASE + notebook first layer seed=0 acc=0.806 f1w=0.810
22
+ [0] BASE + notebook first layer seed=1 acc=0.806 f1w=0.816
23
+ [0] BASE + notebook first layer seed=2 acc=0.811 f1w=0.823
24
+ [0] BASE + notebook first layer seed=3 acc=0.766 f1w=0.776
25
+ [0] BASE + notebook first layer seed=7 acc=0.794 f1w=0.805
26
+ [0] BASE + notebook first layer seed=42 acc=0.794 f1w=0.805
27
+ [0] BASE + notebook first layer seed=123 acc=0.760 f1w=0.773
28
+ [0] BASE + notebook first layer seed=2024 acc=0.789 f1w=0.798
29
+ [0] BASE + notebook first layer seed=5 acc=0.743 f1w=0.754
30
+ [0] BASE + notebook first layer seed=11 acc=0.794 f1w=0.803
31
+ SHARD 0 DONE
diagnostics/2026-11-06-test-split-reproduction/results/BRCA_summary.txt ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+
2
+ === BRCA (paper: F1 0.82)
3
+ seeds accuracy macro F1 weighted F1 precision recall macro F1 min/max collapsed
4
+ config
5
+ BASE + notebook first layer 10 0.786 ± 0.023 0.762 ± 0.027 0.796 ± 0.022 0.773 ± 0.028 0.780 ± 0.038 0.730 / 0.804 0
6
+ BASE lr1e-3 wd1e-4 10 0.748 ± 0.031 0.726 ± 0.032 0.759 ± 0.030 0.739 ± 0.029 0.745 ± 0.045 0.668 / 0.766 0
7
+ D default 10 0.752 ± 0.033 0.720 ± 0.036 0.762 ± 0.032 0.745 ± 0.032 0.742 ± 0.047 0.676 / 0.793 0
diagnostics/2026-11-06-test-split-reproduction/results/COAD.jsonl ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"config": "D default", "seed": 0, "first_layer": "masked_weights", "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 0, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.6730769230769231, "f1_macro": 0.2845528455284553, "f1_weighted": 0.599437148217636, "precision_macro": 0.3421985815602837, "recall_macro": 0.29126984126984123, "epochs_completed": 29, "pred_counts": [47, 3, 0, 2], "collapsed": false, "recall": [0.9428571428571428, 0.2222222222222222, 0.0, 0.0]}
2
+ {"config": "D default", "seed": 1, "first_layer": "masked_weights", "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 1, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.6730769230769231, "f1_macro": 0.20114942528735633, "f1_weighted": 0.541556145004421, "precision_macro": 0.16826923076923078, "recall_macro": 0.25, "epochs_completed": 31, "pred_counts": [52, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0]}
3
+ {"config": "D default", "seed": 2, "first_layer": "masked_weights", "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.6730769230769231, "f1_macro": 0.20114942528735633, "f1_weighted": 0.541556145004421, "precision_macro": 0.16826923076923078, "recall_macro": 0.25, "epochs_completed": 14, "pred_counts": [52, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0]}
4
+ {"config": "D default", "seed": 3, "first_layer": "masked_weights", "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 3, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.6730769230769231, "f1_macro": 0.20114942528735633, "f1_weighted": 0.541556145004421, "precision_macro": 0.16826923076923078, "recall_macro": 0.25, "epochs_completed": 53, "pred_counts": [52, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0]}
5
+ {"config": "D default", "seed": 7, "first_layer": "masked_weights", "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 7, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.6346153846153846, "f1_macro": 0.23123262279888787, "f1_weighted": 0.5456262921508519, "precision_macro": 0.22916666666666666, "recall_macro": 0.2563492063492063, "epochs_completed": 40, "pred_counts": [48, 4, 0, 0], "collapsed": false, "recall": [0.9142857142857143, 0.1111111111111111, 0.0, 0.0]}
6
+ {"config": "D default", "seed": 42, "first_layer": "masked_weights", "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.36538461538461536, "f1_macro": 0.2188825373409552, "f1_weighted": 0.40316174697513446, "precision_macro": 0.25554347826086954, "recall_macro": 0.22619047619047616, "epochs_completed": 38, "pred_counts": [23, 25, 4, 0], "collapsed": false, "recall": [0.42857142857142855, 0.3333333333333333, 0.14285714285714285, 0.0]}
7
+ {"config": "D default", "seed": 123, "first_layer": "masked_weights", "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 123, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.6730769230769231, "f1_macro": 0.20114942528735633, "f1_weighted": 0.541556145004421, "precision_macro": 0.16826923076923078, "recall_macro": 0.25, "epochs_completed": 13, "pred_counts": [52, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0]}
8
+ {"config": "D default", "seed": 2024, "first_layer": "masked_weights", "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2024, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.17307692307692307, "f1_macro": 0.08927738927738926, "f1_weighted": 0.10329926483772636, "precision_macro": 0.16145833333333334, "recall_macro": 0.2642857142857143, "epochs_completed": 30, "pred_counts": [4, 0, 48, 0], "collapsed": false, "recall": [0.05714285714285714, 0.0, 1.0, 0.0]}
9
+ {"config": "D default", "seed": 5, "first_layer": "masked_weights", "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 5, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.5769230769230769, "f1_macro": 0.24386724386724387, "f1_weighted": 0.5369075369075369, "precision_macro": 0.2976190476190476, "recall_macro": 0.24285714285714288, "epochs_completed": 45, "pred_counts": [42, 0, 2, 8], "collapsed": false, "recall": [0.8285714285714286, 0.0, 0.14285714285714285, 0.0]}
10
+ {"config": "D default", "seed": 11, "first_layer": "masked_weights", "hyperparameters": {"epochs": 200, "batch_size": 32, "learning_rate": 0.01, "weight_decay": 0.01, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 11, "biological_activation": "tanh", "projection_activation": "relu", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.6730769230769231, "f1_macro": 0.20114942528735633, "f1_weighted": 0.541556145004421, "precision_macro": 0.16826923076923078, "recall_macro": 0.25, "epochs_completed": 13, "pred_counts": [52, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0]}
11
+ {"config": "BASE lr1e-3 wd1e-4", "seed": 0, "first_layer": "masked_weights", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 0, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.6730769230769231, "f1_macro": 0.20114942528735633, "f1_weighted": 0.541556145004421, "precision_macro": 0.16826923076923078, "recall_macro": 0.25, "epochs_completed": 33, "pred_counts": [52, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0]}
12
+ {"config": "BASE lr1e-3 wd1e-4", "seed": 1, "first_layer": "masked_weights", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 1, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.6730769230769231, "f1_macro": 0.20114942528735633, "f1_weighted": 0.541556145004421, "precision_macro": 0.16826923076923078, "recall_macro": 0.25, "epochs_completed": 20, "pred_counts": [52, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0]}
13
+ {"config": "BASE lr1e-3 wd1e-4", "seed": 2, "first_layer": "masked_weights", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.6730769230769231, "f1_macro": 0.20114942528735633, "f1_weighted": 0.541556145004421, "precision_macro": 0.16826923076923078, "recall_macro": 0.25, "epochs_completed": 14, "pred_counts": [52, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0]}
14
+ {"config": "BASE lr1e-3 wd1e-4", "seed": 3, "first_layer": "masked_weights", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 3, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.6730769230769231, "f1_macro": 0.20114942528735633, "f1_weighted": 0.541556145004421, "precision_macro": 0.16826923076923078, "recall_macro": 0.25, "epochs_completed": 23, "pred_counts": [52, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0]}
15
+ {"config": "BASE lr1e-3 wd1e-4", "seed": 7, "first_layer": "masked_weights", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 7, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.6730769230769231, "f1_macro": 0.20114942528735633, "f1_weighted": 0.541556145004421, "precision_macro": 0.16826923076923078, "recall_macro": 0.25, "epochs_completed": 40, "pred_counts": [52, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0]}
16
+ {"config": "BASE lr1e-3 wd1e-4", "seed": 42, "first_layer": "masked_weights", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.6730769230769231, "f1_macro": 0.20114942528735633, "f1_weighted": 0.541556145004421, "precision_macro": 0.16826923076923078, "recall_macro": 0.25, "epochs_completed": 26, "pred_counts": [52, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0]}
17
+ {"config": "BASE lr1e-3 wd1e-4", "seed": 123, "first_layer": "masked_weights", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 123, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.6730769230769231, "f1_macro": 0.20114942528735633, "f1_weighted": 0.541556145004421, "precision_macro": 0.16826923076923078, "recall_macro": 0.25, "epochs_completed": 15, "pred_counts": [52, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0]}
18
+ {"config": "BASE lr1e-3 wd1e-4", "seed": 2024, "first_layer": "masked_weights", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2024, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.6730769230769231, "f1_macro": 0.20114942528735633, "f1_weighted": 0.541556145004421, "precision_macro": 0.16826923076923078, "recall_macro": 0.25, "epochs_completed": 27, "pred_counts": [52, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0]}
19
+ {"config": "BASE lr1e-3 wd1e-4", "seed": 5, "first_layer": "masked_weights", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 5, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.6730769230769231, "f1_macro": 0.20114942528735633, "f1_weighted": 0.541556145004421, "precision_macro": 0.16826923076923078, "recall_macro": 0.25, "epochs_completed": 44, "pred_counts": [52, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0]}
20
+ {"config": "BASE lr1e-3 wd1e-4", "seed": 11, "first_layer": "masked_weights", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 11, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.6730769230769231, "f1_macro": 0.20114942528735633, "f1_weighted": 0.541556145004421, "precision_macro": 0.16826923076923078, "recall_macro": 0.25, "epochs_completed": 36, "pred_counts": [52, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0]}
21
+ {"config": "BASE + notebook first layer", "seed": 0, "first_layer": "dense_then_mask", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 0, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.5384615384615384, "f1_macro": 0.22095238095238096, "f1_weighted": 0.49963369963369964, "precision_macro": 0.20416666666666666, "recall_macro": 0.24126984126984127, "epochs_completed": 18, "pred_counts": [40, 12, 0, 0], "collapsed": false, "recall": [0.7428571428571429, 0.2222222222222222, 0.0, 0.0]}
22
+ {"config": "BASE + notebook first layer", "seed": 1, "first_layer": "dense_then_mask", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 1, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.6538461538461539, "f1_macro": 0.244249726177437, "f1_weighted": 0.5666863257224704, "precision_macro": 0.296875, "recall_macro": 0.2634920634920635, "epochs_completed": 20, "pred_counts": [48, 2, 2, 0], "collapsed": false, "recall": [0.9428571428571428, 0.1111111111111111, 0.0, 0.0]}
23
+ {"config": "BASE + notebook first layer", "seed": 2, "first_layer": "dense_then_mask", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.6730769230769231, "f1_macro": 0.20114942528735633, "f1_weighted": 0.541556145004421, "precision_macro": 0.16826923076923078, "recall_macro": 0.25, "epochs_completed": 14, "pred_counts": [52, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0]}
24
+ {"config": "BASE + notebook first layer", "seed": 3, "first_layer": "dense_then_mask", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 3, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.6730769230769231, "f1_macro": 0.20114942528735633, "f1_weighted": 0.541556145004421, "precision_macro": 0.16826923076923078, "recall_macro": 0.25, "epochs_completed": 13, "pred_counts": [52, 0, 0, 0], "collapsed": true, "recall": [1.0, 0.0, 0.0, 0.0]}
25
+ {"config": "BASE + notebook first layer", "seed": 7, "first_layer": "dense_then_mask", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 7, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.4807692307692308, "f1_macro": 0.18777777777777777, "f1_weighted": 0.45000000000000007, "precision_macro": 0.17777777777777776, "recall_macro": 0.19920634920634922, "epochs_completed": 28, "pred_counts": [40, 9, 3, 0], "collapsed": false, "recall": [0.6857142857142857, 0.1111111111111111, 0.0, 0.0]}
26
+ {"config": "BASE + notebook first layer", "seed": 42, "first_layer": "dense_then_mask", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 42, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.4807692307692308, "f1_macro": 0.16447368421052633, "f1_weighted": 0.44281376518218624, "precision_macro": 0.1524390243902439, "recall_macro": 0.17857142857142858, "epochs_completed": 25, "pred_counts": [41, 4, 7, 0], "collapsed": false, "recall": [0.7142857142857143, 0.0, 0.0, 0.0]}
27
+ {"config": "BASE + notebook first layer", "seed": 123, "first_layer": "dense_then_mask", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 123, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.6346153846153846, "f1_macro": 0.23123262279888787, "f1_weighted": 0.5456262921508519, "precision_macro": 0.22916666666666666, "recall_macro": 0.2563492063492063, "epochs_completed": 22, "pred_counts": [48, 4, 0, 0], "collapsed": false, "recall": [0.9142857142857143, 0.1111111111111111, 0.0, 0.0]}
28
+ {"config": "BASE + notebook first layer", "seed": 2024, "first_layer": "dense_then_mask", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 2024, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.5576923076923077, "f1_macro": 0.20889894419306185, "f1_weighted": 0.503596704954171, "precision_macro": 0.19404069767441862, "recall_macro": 0.2277777777777778, "epochs_completed": 23, "pred_counts": [43, 8, 1, 0], "collapsed": false, "recall": [0.8, 0.1111111111111111, 0.0, 0.0]}
29
+ {"config": "BASE + notebook first layer", "seed": 5, "first_layer": "dense_then_mask", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 5, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.5, "f1_macro": 0.17333333333333334, "f1_weighted": 0.4666666666666667, "precision_macro": 0.1625, "recall_macro": 0.18571428571428572, "epochs_completed": 24, "pred_counts": [40, 7, 5, 0], "collapsed": false, "recall": [0.7428571428571429, 0.0, 0.0, 0.0]}
30
+ {"config": "BASE + notebook first layer", "seed": 11, "first_layer": "dense_then_mask", "hyperparameters": {"epochs": 100, "batch_size": 16, "learning_rate": 0.001, "weight_decay": 0.0001, "dropout": 0.3, "projection_dim": 64, "fusion_dim": 12, "test_fraction": 0.2, "early_stopping_fraction": 0.125, "optimizer": "Adam", "class_weighting": true, "early_stopping_patience": 12, "seed": 11, "biological_activation": "relu", "projection_activation": "sigmoid", "fusion_activation": "tanh", "sgd_momentum": 0.03}, "split": "test", "test_samples": 52, "accuracy": 0.5769230769230769, "f1_macro": 0.23444976076555024, "f1_weighted": 0.5259477364740524, "precision_macro": 0.21666666666666667, "recall_macro": 0.25555555555555554, "epochs_completed": 26, "pred_counts": [42, 10, 0, 0], "collapsed": false, "recall": [0.8, 0.2222222222222222, 0.0, 0.0]}
diagnostics/2026-11-06-test-split-reproduction/results/COAD_shard0.log ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [0] D default seed=0 acc=0.673 f1w=0.599
2
+ [0] D default seed=1 acc=0.673 f1w=0.542
3
+ [0] D default seed=2 acc=0.673 f1w=0.542
4
+ [0] D default seed=3 acc=0.673 f1w=0.542
5
+ [0] D default seed=7 acc=0.635 f1w=0.546
6
+ [0] D default seed=42 acc=0.365 f1w=0.403
7
+ [0] D default seed=123 acc=0.673 f1w=0.542
8
+ [0] D default seed=2024 acc=0.173 f1w=0.103
9
+ [0] D default seed=5 acc=0.577 f1w=0.537
10
+ [0] D default seed=11 acc=0.673 f1w=0.542
11
+ [0] BASE lr1e-3 wd1e-4 seed=0 acc=0.673 f1w=0.542
12
+ [0] BASE lr1e-3 wd1e-4 seed=1 acc=0.673 f1w=0.542
13
+ [0] BASE lr1e-3 wd1e-4 seed=2 acc=0.673 f1w=0.542
14
+ [0] BASE lr1e-3 wd1e-4 seed=3 acc=0.673 f1w=0.542
15
+ [0] BASE lr1e-3 wd1e-4 seed=7 acc=0.673 f1w=0.542
16
+ [0] BASE lr1e-3 wd1e-4 seed=42 acc=0.673 f1w=0.542
17
+ [0] BASE lr1e-3 wd1e-4 seed=123 acc=0.673 f1w=0.542
18
+ [0] BASE lr1e-3 wd1e-4 seed=2024 acc=0.673 f1w=0.542
19
+ [0] BASE lr1e-3 wd1e-4 seed=5 acc=0.673 f1w=0.542
20
+ [0] BASE lr1e-3 wd1e-4 seed=11 acc=0.673 f1w=0.542
21
+ [0] BASE + notebook first layer seed=0 acc=0.538 f1w=0.500
22
+ [0] BASE + notebook first layer seed=1 acc=0.654 f1w=0.567
23
+ [0] BASE + notebook first layer seed=2 acc=0.673 f1w=0.542
24
+ [0] BASE + notebook first layer seed=3 acc=0.673 f1w=0.542
25
+ [0] BASE + notebook first layer seed=7 acc=0.481 f1w=0.450
26
+ [0] BASE + notebook first layer seed=42 acc=0.481 f1w=0.443
27
+ [0] BASE + notebook first layer seed=123 acc=0.635 f1w=0.546
28
+ [0] BASE + notebook first layer seed=2024 acc=0.558 f1w=0.504
29
+ [0] BASE + notebook first layer seed=5 acc=0.500 f1w=0.467
30
+ [0] BASE + notebook first layer seed=11 acc=0.577 f1w=0.526
31
+ SHARD 0 DONE