scripts/aggregate_ablation_results.py

from __future__ import annotations

import argparse
import json
from pathlib import Path

import pandas as pd

from common import ROOT


METRICS = ["accuracy", "precision", "recall", "f1", "macro_f1"]
VARIANT_NAMES = {
    "dynafall_joint": "A1 joint",
    "dynafall_joint_bone": "A2 joint+bone",
    "dynafall_full_no_dropout": "A3 +dynamics",
    "dynafall_random_dropout": "A4 +random dropout",
    "dynafall": "A5 +confidence dropout",
}


def main() -> None:
    ap = argparse.ArgumentParser()
    ap.add_argument("--root", default="results/ablation")
    ap.add_argument("--out", default="results/ablation_summary.csv")
    args = ap.parse_args()

    rows = []
    for path in (ROOT / args.root).glob("seed_*/*/*/metrics_test_clean.json"):
        row = json.loads(path.read_text())
        row["seed"] = int(path.parts[-4].replace("seed_", ""))
        row["variant"] = VARIANT_NAMES.get(row["method"], row["method"])
        rows.append(row)
    if not rows:
        raise SystemExit(f"No ablation metrics found under {args.root}")

    df = pd.DataFrame(rows)
    out = ROOT / args.out
    out.parent.mkdir(parents=True, exist_ok=True)
    df.to_csv(out.with_name(out.stem + "_raw.csv"), index=False)
    summary = df.groupby(["dataset", "variant"], as_index=False)[METRICS].agg(["mean", "std"])
    summary.to_csv(out)
    print(f"Wrote {out}")

    flat = []
    order = {v: i for i, v in enumerate(VARIANT_NAMES.values())}
    for (dataset, variant), group in df.groupby(["dataset", "variant"]):
        row = {"dataset": dataset, "variant": variant, "n": len(group)}
        for metric in METRICS:
            row[metric] = f"{group[metric].mean():.4f} +/- {group[metric].std(ddof=1):.4f}"
        flat.append(row)
    flat_df = pd.DataFrame(flat)
    flat_df["order"] = flat_df["variant"].map(order)
    flat_df = flat_df.sort_values(["dataset", "order"]).drop(columns=["order"])
    print(flat_df.to_string(index=False))


if __name__ == "__main__":
    main()