Эксперимент: Excitation-Controlled Recurrent Learning (#828)

{ "worked": false, "confidence": 9, "verdict": "Built a reproducible linear recurrent identification MVP with empirical feature Gramians, greedy pulse selection, parameter recovery, and fixed-seed sweeps. The mechanism checks passed: Gramian minimum eigenvalue scaled as amplitude squared, vanished at an unexcited channel, and approximately followed b squared scaling. However, under equal input energy the designed pulse set had lower lambda_min than passive excitation (0.05425 vs 0.07168), so the central promised Gramian improvement was not observed; active data nevertheless reduced parameter error.", "metrics": { "baseline": "Equal-energy passive trajectories: lambda_min(G)=0.071684; parameter error 1.638 at noise 0.02 and 1.637 at noise 0.1.", "idea": "Greedy equal-energy bounded pulses: lambda_min(G)=0.054250, ratio 0.757; parameter error 0.630 at noise 0.02 and 0.634 at noise 0.1. Amplitude sweep observed normalized [0.0625000000, 0.2500000000, 1, 4.0000000000] versus predicted [0.0625, 0.25, 1, 4]." }, "how_to_run": "/home/maxwelhelp/main/bin/python3 experiment.py", "files": [ "experiment.py", "results.json" ], "limitations": "This is a 2D linear recurrent toy, not the proposed 20D sparse tanh RNN with five observed coordinates. No online neural autodiff Jacobian estimation, latent interventions, sparse loss, long-horizon prediction, FLOP accounting, or robustness to model misspecification was tested." }