Datasets:
File size: 4,208 Bytes
84a242a | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 | #!/usr/bin/env python3
"""Aggregate all fold/seed/checkpoint metrics for paper analysis."""
import json
import argparse
from pathlib import Path
import pandas as pd
from run_experiment import METHODS, atomic_json
from release_support import add_path_arguments
def main() -> None:
parser = argparse.ArgumentParser(description="Summarize existing acquisition diagnostics; no MolAS training or figure reproduction.")
add_path_arguments(parser, inputs=False)
parser.add_argument("--methods", nargs="+", choices=METHODS, default=list(METHODS))
parser.add_argument("--folds", nargs="+", type=int, choices=range(5), default=list(range(5)))
parser.add_argument("--seeds", nargs="+", type=int, default=[0, 1, 2])
parser.add_argument("--smoke-test", action="store_true")
args = parser.parse_args()
if "plain_mean_ucb" not in args.methods:
parser.error("The historical ratio/delta summaries require plain_mean_ucb in --methods")
records = []
shares = []
for method in args.methods:
for fold in args.folds:
for seed in args.seeds:
root = args.output_dir / ("smoke" if args.smoke_test else "runs") / method / f"fold{fold}" / f"seed{seed}"
for percentage in range(1, 101):
label = f"{percentage:03d}pct"
reconstruction = json.loads((root / "reconstruction_metrics" / f"metrics_{label}.json").read_text())
runtime = json.loads((root / "runtime_prediction_metrics" / f"metrics_{label}.json").read_text())
records.append({"method": method, "fold": fold, "seed": seed, **reconstruction, **runtime})
if percentage in {10, 20, 30, 40}:
distribution = pd.read_csv(root / "algorithm_query_distributions" / f"distribution_{label}.csv")
shares.append({
"method": method, "fold": fold, "seed": seed,
"percentage": percentage,
"max_algorithm_share": distribution["cumulative_share"].max(),
"min_algorithm_share": distribution["cumulative_share"].min(),
})
frame = pd.DataFrame(records)
summary = frame.groupby(["method", "percentage"]).agg(
runtime_mean=("cumulative_true_runtime", "mean"),
runtime_std=("cumulative_true_runtime", "std"),
unqueried_mae_mean=("unqueried_mae", "mean"),
unqueried_mae_std=("unqueried_mae", "std"),
ndcg_at_3_mean=("ndcg_at_3", "mean"),
ndcg_at_3_std=("ndcg_at_3", "std"),
top1_mean=("top1_agreement", "mean"),
top1_std=("top1_agreement", "std"),
vbs_regret_mean=("vbs_regret", "mean"),
vbs_regret_std=("vbs_regret", "std"),
).reset_index()
baseline = summary[summary.method == "plain_mean_ucb"].set_index("percentage")
summary["runtime_ratio_to_plain"] = summary.apply(
lambda row: row.runtime_mean / baseline.loc[row.percentage, "runtime_mean"], axis=1
)
summary["mae_delta_to_plain"] = summary.apply(
lambda row: row.unqueried_mae_mean - baseline.loc[row.percentage, "unqueried_mae_mean"], axis=1
)
summary["ndcg_delta_to_plain"] = summary.apply(
lambda row: row.ndcg_at_3_mean - baseline.loc[row.percentage, "ndcg_at_3_mean"], axis=1
)
output = args.output_dir / ("smoke_summary" if args.smoke_test else "summary")
output.mkdir(parents=True, exist_ok=True)
atomic_json(output / "summary_scope.json", {"methods": args.methods, "folds": args.folds,
"seeds": args.seeds, "smoke_test": args.smoke_test, "records": len(frame),
"aggregation": "Descriptive means and standard deviations over the selected fold-seed records; not independent confidence intervals."})
frame.to_csv(output / "all_checkpoint_metrics.csv", index=False)
summary.to_csv(output / "method_checkpoint_summary.csv", index=False)
pd.DataFrame(shares).to_csv(output / "algorithm_share_key_checkpoints.csv", index=False)
print(summary[summary.percentage.isin([10, 20, 30, 40])].to_string(index=False))
if __name__ == "__main__":
main()
|