Эксперимент: Pick-to-Learn Scenario Compression for Safe NN Calibration (#412)

{ "worked": true, "confidence": 8, "verdict": "Built a readable Pick-to-Learn MVP with exact theorem-bound evaluation, greedy largest-violation compression, constrained calibration, and iid fresh-scenario diagnostics. The mechanism manifested: epsilon decreased roughly by half when N doubled, increased with weaker beta and larger k, and greedy k=2 achieved 0.774% fresh violation versus 54.162% for random k=2 in the toy problem. The paper’s displayed equation gives 3.53% for N=400,k=2,beta=1e-5, differing from its quoted 4.8%; this implementation reports the literal equation result rather than silently using 4.8%.", "metrics": { "baseline": "Random compression: fresh violation 54.162% (k=2), 36.594% (k=4), 29.870% (k=8); theorem epsilon at N=400,k=2,beta=1e-5 is 3.530%.", "idea": "Greedy compression: fresh violation 0.774% (k=2), 0.394% (k=4), 0.394% (k=8); theorem epsilon sweep is 13.549%, 6.963%, 3.530%, 1.777% for N=100,200,400,800 and rises monotonically with k." }, "how_to_run": "/home/maxwelhelp/main/bin/python3 p2l_experiment.py", "files": [ "p2l_experiment.py", "results.json", "run_output.txt", "REPORT.txt" ], "limitations": "The benchmark is a two-parameter affine convex surrogate, not an MLP on MNIST-C, MPC, or rollout calibration. Only one seed and one random-compression draw were used; no formal theorem claim is made for adaptive neural-network training, and the shifted-distribution diagnostic is not a certificate test." }