5xuekun's picture
SHAPE run logs: first batch (traces, scores, embeddings, checkpoints; no image files)
2fa4eed
Raw History Blame Contribute Delete
5.65 kB
{
"primary": "day_classifier",
"day_classifier": {
"target": "time point (day), eight classes",
"classes": [
0,
2,
4,
6,
8,
10,
12,
16
],
"C": 0.01,
"trials": [
{
"C": 0.01,
"validation": {
"accuracy": 0.35294117647058826,
"balanced_accuracy": 0.35625,
"n": 34
}
},
{
"C": 0.1,
"validation": {
"accuracy": 0.3235294117647059,
"balanced_accuracy": 0.325,
"n": 34
}
},
{
"C": 1.0,
"validation": {
"accuracy": 0.3235294117647059,
"balanced_accuracy": 0.325,
"n": 34
}
},
{
"C": 10.0,
"validation": {
"accuracy": 0.3235294117647059,
"balanced_accuracy": 0.325,
"n": 34
}
}
],
"selection": "development_val balanced accuracy",
"fit_split": "development_train",
"selection_split": "development_val",
"refit_on_validation": false,
"train_crops": 88,
"validation_crops": 34,
"training_wells": [
"w00",
"w01",
"w03",
"w04",
"w06",
"w10",
"w11",
"w12",
"w14",
"w17",
"w19"
],
"training_weighting": "equal total weight per training well",
"features": "row L2 norm, train-fitted standard scaling",
"classes_seen_in_training": [
0,
2,
4,
6,
8,
10,
12,
16
]
},
"day_regression": {
"target": "day",
"alpha": 100.0,
"trials": [
{
"alpha": 0.1,
"validation_mae_days": 2.6218900132923797
},
{
"alpha": 1.0,
"validation_mae_days": 2.620826152515704
},
{
"alpha": 10.0,
"validation_mae_days": 2.6110048194582127
},
{
"alpha": 100.0,
"validation_mae_days": 2.5595123095855272
}
],
"selection": "development_val MAE",
"fit_split": "development_train",
"refit_on_validation": false
},
"condition_probe": {
"target": "condition == FC",
"C": 1.0,
"trials": [
{
"C": 0.01,
"validation": {
"n": 34,
"positives": 18,
"accuracy": 0.7941176470588235,
"balanced_accuracy": 0.7916666666666667,
"auroc": 0.9027777777777778,
"sensitivity": 0.8333333333333334,
"specificity": 0.75,
"confusion": {
"tp": 15,
"tn": 12,
"fp": 4,
"fn": 3
}
}
},
{
"C": 0.1,
"validation": {
"n": 34,
"positives": 18,
"accuracy": 0.8235294117647058,
"balanced_accuracy": 0.8194444444444444,
"auroc": 0.8993055555555556,
"sensitivity": 0.8888888888888888,
"specificity": 0.75,
"confusion": {
"tp": 16,
"tn": 12,
"fp": 4,
"fn": 2
}
}
},
{
"C": 1.0,
"validation": {
"n": 34,
"positives": 18,
"accuracy": 0.8529411764705882,
"balanced_accuracy": 0.8506944444444444,
"auroc": 0.8958333333333333,
"sensitivity": 0.8888888888888888,
"specificity": 0.8125,
"confusion": {
"tp": 16,
"tn": 13,
"fp": 3,
"fn": 2
}
}
},
{
"C": 10.0,
"validation": {
"n": 34,
"positives": 18,
"accuracy": 0.8235294117647058,
"balanced_accuracy": 0.8194444444444444,
"auroc": 0.9201388888888888,
"sensitivity": 0.8888888888888888,
"specificity": 0.75,
"confusion": {
"tp": 16,
"tn": 12,
"fp": 4,
"fn": 2
}
}
}
],
"classifier_probability_threshold": 0.5,
"selection": "development_val balanced accuracy",
"fit_split": "development_train",
"selection_split": "development_val",
"refit_on_validation": false,
"train_crops": 88,
"validation_crops": 34,
"training_wells": [
"w00",
"w01",
"w03",
"w04",
"w06",
"w10",
"w11",
"w12",
"w14",
"w17",
"w19"
],
"training_weighting": "equal total weight per training well",
"features": "row L2 norm, train-fitted standard scaling",
"weights_sha256": "8bc28930b9f032fc318d51a08fbffc8c7103037b17f0c7c198544c5fdb5a7142"
},
"gfp_probe": {
"target": "gfp_integrated_bgsub",
"alpha": 100.0,
"trials": [
{
"alpha": 0.1,
"validation_mae": 179647817.94266015
},
{
"alpha": 1.0,
"validation_mae": 179350034.7235453
},
{
"alpha": 10.0,
"validation_mae": 176629318.39557275
},
{
"alpha": 100.0,
"validation_mae": 160977018.56452122
}
],
"selection": "development_val MAE",
"fit_split": "development_train",
"refit_on_validation": false,
"target_standardisation": {
"mean": 140753762.16472122,
"sd": 189465895.65136117
}
},
"gfp_label_baseline": {
"target": "gfp_integrated_bgsub",
"design": "intercept + one-hot day (reference day 0) + condition",
"fit_split": "development_train",
"n_parameters": 9
},
"input_embedding_sha256": "bbefd10f6a530ae507f29c5d267eeb02e617364280beb701f7f8b9a603352cb9",
"prediction_channels": [
"PhaseContrast"
]
}