# Эксперимент: Uncertainty-guided family sampling (#400)

{
  "worked": true,
  "confidence": 8,
  "verdict": "Built and numerically tested an uncertainty-guided family allocation controller with exploration mixing. The variance identity matched Monte Carlo within 2.3%, allocation proportional to u reduced predicted variance by a stable 1.46x across budgets, and exploration prevented severe mode collapse when one high-variance family was deliberately mispredicted. The controller achieved predicted/empirical variance 0.899/0.893 versus 1.258/1.291 for uniform and 0.913/0.917 for magnitude-only sampling, though its gain over magnitude-only was modest.",
  "metrics": {
    "baseline": "Uniform: predicted variance 1.2576, empirical variance 1.2911; magnitude-only: predicted 0.9129, empirical 0.9170.",
    "idea": "Uncertainty-guided controller: predicted variance 0.8986, empirical variance 0.8928; proportional-allocation sweep gave 1.4628x uniform variance reduction at budget 4096. Under hidden high-uncertainty family, epsilon 0.0/0.01/0.05/0.2 yielded family counts 3/5/10/33 and true predicted variances 8.97/5.68/3.22/1.51."
  },
  "how_to_run": "/home/maxwelhelp/main/bin/python3 experiment.py",
  "files": [
    "experiment.py",
    "results.json",
    "run.txt"
  ],
  "limitations": "This is a synthetic independent-Gaussian family model with known uncertainties, not a neural predictor or real permutation-family Monte Carlo system. It does not test learned uncertainty calibration, importance-weight correction, correlated family noise, estimator bias, wall-clock efficiency, or a full training/active-learning loop."
}