# Эксперимент: Controlled Stationary Hyperparameter Sweep (#868)

{
  "worked": false,
  "confidence": 9,
  "verdict": "Built an analytic controlled SGLD sweep for the Gaussian family rho(x|a)=N(0,1/a), including covariance-response estimation, continuity-equation verification, particle sweeps, and a deterministic moment recurrence. The covariance identity matched -1/a^2 closely and the continuity residual was 1.36e-17; uncontrolled bias scaled approximately linearly with sweep rate (slope 0.90), while control reduced fastest-rate RMS error by 189x. However, the controlled deterministic error scaled with slope 0.84 rather than the predicted quadratic slope, so the required mechanism-level prediction was not confirmed under the tested discretization and worked=false.",
  "metrics": {
    "baseline": "Uncontrolled deterministic RMS tracking error: 0.01008, 0.01962, 0.03740, 0.06912, 0.12260 for rates 0.01, 0.02, 0.04, 0.08, 0.16; log-log slope 0.903.",
    "idea": "Controlled deterministic RMS error: 6.35e-05, 1.22e-04, 2.27e-04, 3.997e-04, 6.478e-04; log-log slope 0.842 and fastest-rate ratio 0.00528 versus baseline. Covariance responses at a=0.5,1,2 were -3.987,-1.0003,-0.2453 versus exact -4,-1,-0.25."
  },
  "how_to_run": "/home/maxwelhelp/main/bin/python3 controlled_sweep_experiment.py",
  "files": [
    "controlled_sweep_experiment.py",
    "results.json"
  ],
  "limitations": "Only a one-dimensional Gaussian toy family was tested, with an analytically known control field and fixed-step frozen-coefficient OU integration. No learned control network, unknown-score setting, neural-network weights, MNIST, FLOP/speed comparison, or independent-equilibration baseline was evaluated; the controlled O(v^2) asymptotic was not isolated by a timestep-convergence study."
}