Эксперимент: Hankel Residual Observer (#1363)

{ "worked": true, "confidence": 7, "verdict": "Built a bounded Hankel residual observer with ridge reconstruction and confidence-scaled correction around a small GRU forecaster. Exact period-4 residuals were reconstructed with RMSE 8.6e-13, and disturbed-stream RMSE improved from 0.9907 to 0.8792. Clean-stream RMSE worsened from 0.05314 to 0.05495, so the benefit is promising but not robust across regimes.", "metrics": { "baseline": "Disturbed RMSE 0.9907; disturbed long-horizon RMSE 1.1564; clean RMSE 0.05314; clean long-horizon RMSE 0.05773.", "idea": "Disturbed RMSE 0.8792; disturbed long-horizon RMSE 1.1278; clean RMSE 0.05495; clean long-horizon RMSE 0.06024; p95 observer latency 107.3 ms disturbed and 95.7 ms clean." }, "how_to_run": "/home/maxwelhelp/main/bin/python3 run_experiment.py", "files": [ "run_experiment.py" ], "limitations": "Only a synthetic AR(1) plus sinusoidal-disturbance stream was tested. No real benchmark, validation-based hyperparameter selection, multi-seed uncertainty, GPU timing, baseline latency, memory, or optimized incremental Gram updates were evaluated." }