Эксперимент: Prescribed-Performance Hidden-State Observer (#928)

{ "worked": false, "confidence": 9, "verdict": "Built a NumPy PPO observer with shrinking prescribed envelope, transformed innovation, Euler updates, gain sweep, and a small noisy-sinusoid GRU comparison. The toy check clearly reproduced the claimed behavior: modest gains kept normalized innovation below 0.18, while gains at least 5 reached the clamp and generated large estimates. However, PPO-corrected GRU RMSE was 1.672 versus 0.274 raw GRU and 0.221 for the fixed observer, so no predictive win was observed.", "metrics": { "baseline": "Raw GRU RMSE 0.2743; fixed-gain observer RMSE 0.2207; toy fixed observer max normalized innovation 0.1764", "idea": "PPO toy max normalized innovation 0.0526 with 0 violations; gain sweep reached |xi|=0.999999 for gains >=5; GRU-corrected RMSE 1.6721, mean |xi| 0.8831, 0 envelope violations" }, "how_to_run": "/home/maxwelhelp/main/bin/python3 ppo_observer_experiment.py", "files": [ "ppo_observer_experiment.py", "results.json" ], "limitations": "Only one fixed random seed, one noisy sinusoid task, one small GRU, and a limited gain sweep were tested. The implementation uses an output-residual correction rather than training the PPO jointly as an auxiliary loss, and no Mackey-Glass, long-horizon rollout, FLOP, or multi-seed study was performed." }