# Эксперимент: Tail-triggered adaptive ridge head (#1423)

{
  "worked": false,
  "confidence": 9,
  "verdict": "Built the spectral inverse-trace/hard-edge diagnostics, closed-form ridge head, adaptive controller, and fixed-seed synthetic regression experiment. The decomposition and ridge normal equations were verified to numerical precision, but the promised performance effect was not observed: adaptive ridge saturated at gamma=0.05 and had substantially worse hard-edge MSE (0.0160) and p99 loss (0.1022) than fixed ridge (0.0128, 0.0831) and ridgeless regression (0.0135, 0.0875).",
  "metrics": {
    "baseline": "Ridgeless hard-edge MSE 0.01350, p99 0.08750; fixed ridge hard-edge MSE 0.01281, p99 0.08314.",
    "idea": "Adaptive hard-edge MSE 0.01602, p99 0.10216, median gamma 0.05 at the upper clip; healthy MSE 0.01338 and p99 0.08562. Math split absolute error 2.27e-13 and ridge normal-equation residual 9.15e-17."
  },
  "how_to_run": "/home/maxwelhelp/main/bin/python3 adaptive_ridge_experiment.py",
  "files": [
    "adaptive_ridge_experiment.py",
    "results.json"
  ],
  "limitations": "This is a small synthetic linear-head test, not a neural backbone or CIFAR-10 experiment. It uses one-batch controller updates and does not test multi-step EMA dynamics, corruption distributions beyond the constructed hard-edge covariance, throughput, or end-to-end AdamW training."
}