Эксперимент: Support-Budgeted Hamming Polynomial Layer (#1304)

{ "worked": false, "confidence": 9, "verdict": "Built a q=2 Fourier interaction layer with configurable support order, order-wise Bohnenblust–Hille quasi-norm penalty, exact finite-group sanity checks, and matched-parameter parity benchmarks. The coefficient quasi-norm to sup-norm ratios stayed approximately dimension-stable for N=3–8, and the penalty consistently improved Fourier generalization over the unpenalized Fourier model. However, the budgeted layer did not beat the parameter-matched MLP: on 2-way parity it averaged 0.887 validation accuracy versus 1.000 for the MLP, and on 3-way parity it averaged 0.585 versus 0.878, so the claimed overall win was not observed.", "metrics": { "baseline": "Matched MLP: 1.000 mean test accuracy on 2-way parity; 0.878 mean on 3-way parity.", "idea": "BH-penalized Fourier: 0.887 mean test accuracy on 2-way parity and 0.585 mean on 3-way parity; it improved over unpenalized Fourier (0.686 and 0.477 respectively) but remained below the MLP. Math-check order-2 quasi-norm/sup-norm ratios had medians approximately 0.74–0.83 across N=3–8 and maxima approximately 0.91–1.00." }, "how_to_run": "/home/maxwelhelp/main/bin/python3 experiment.py", "files": [ "experiment.py" ], "limitations": "Only q=2 synthetic parity tasks were tested; no q>2 categorical benchmark, real tabular data, S=3 unrestricted-versus-budgeted support ablation, FLOP/runtime measurement, coefficient clipping, hashed sampling, or systematic hyperparameter tuning was performed. The implementation uses real {-1,+1} characters and a scalar output rather than the full complex q-ary multi-output layer." }