Uniform-Certificate Bayesian Feature Head / results.json

✓✓ Beats tuned baseline

Raw ⬇ ZIP
  1{
  2  "seed": 2399,
  3  "identity_check": {
  4    "max_abs_V_difference": 1.9895196601282805e-13,
  5    "max_abs_b_difference": 0.0,
  6    "max_abs_theta_difference": 1.1102230246251565e-15
  7  },
  8  "contraction_sweep": {
  9    "n": [
 10      1,
 11      2,
 12      4,
 13      8,
 14      16,
 15      32,
 16      64,
 17      128,
 18      256,
 19      512,
 20      1024
 21    ],
 22    "sd": [
 23      0.11993338883749853,
 24      0.08482925333272914,
 25      0.059991668402376,
 26      0.04242346089980638,
 27      0.02999895838758367,
 28      0.02121283516040503,
 29      0.01499986979336206,
 30      0.010606555682500188,
 31      0.007499983724011317,
 32      0.005303295104458757,
 33      0.0037499979654964486
 34    ],
 35    "loglog_slope": -0.4999798865649055,
 36    "predicted_slope": -0.5,
 37    "covered_sd_n1024": 0.0037499979654964486,
 38    "uncovered_sd_n1024": 3.5988246905709973,
 39    "uncovered_to_covered": 959.6871048154187
 40  },
 41  "coverage_sweep": [
 42    {
 43      "n": 40,
 44      "region": "concentrated",
 45      "simultaneous_coverage": 0.82,
 46      "mean_interval_width": 11.116197664400383,
 47      "grid_rmse": 0.7736428120455685,
 48      "target_coverage": 0.95,
 49      "beta": 2.5
 50    },
 51    {
 52      "n": 160,
 53      "region": "concentrated",
 54      "simultaneous_coverage": 0.81,
 55      "mean_interval_width": 10.619157921526085,
 56      "grid_rmse": 0.7687284286131765,
 57      "target_coverage": 0.95,
 58      "beta": 2.5
 59    },
 60    {
 61      "n": 40,
 62      "region": "broad",
 63      "simultaneous_coverage": 0.65,
 64      "mean_interval_width": 0.6031675060514103,
 65      "grid_rmse": 0.12698464832956854,
 66      "target_coverage": 0.95,
 67      "beta": 2.5
 68    },
 69    {
 70      "n": 160,
 71      "region": "broad",
 72      "simultaneous_coverage": 0.72,
 73      "mean_interval_width": 0.20452947602262125,
 74      "grid_rmse": 0.03973740651968079,
 75      "target_coverage": 0.95,
 76      "beta": 2.5
 77    }
 78  ],
 79  "baseline_comparison": {
 80    "deterministic_ridge_grid_rmse": 0.6343544717217113,
 81    "certificate_grid_rmse": 0.6343544717217113,
 82    "deterministic_interval_width": 0.0,
 83    "certificate_mean_interval_width": 10.58485278820481,
 84    "certificate_grid_miss_fraction": 0.0,
 85    "feature_dim": 17
 86  },
 87  "parameter_scaling_sweeps": {
 88    "sigma": [
 89      0.03,
 90      0.06,
 91      0.12,
 92      0.24
 93    ],
 94    "mean_sd": [
 95      0.007953485678409046,
 96      0.01590697135681809,
 97      0.03181394271363618,
 98      0.06362788542727237
 99    ],
100    "observed_sigma_ratios": [
101      0.25,
102      0.5,
103      1.0,
104      2.0
105    ],
106    "predicted_sigma_ratios": [
107      0.25,
108      0.5,
109      1.0,
110      2.0
111    ],
112    "lambda": [
113      0.001,
114      0.01,
115      0.1,
116      1.0
117    ],
118    "mean_sd_by_lambda": [
119      0.032761851767028546,
120      0.03276052563108217,
121      0.03274727398347355,
122      0.032615720164693364
123    ],
124    "grid_rmse_by_lambda": [
125      0.024403537650630847,
126      0.037223634767357655,
127      0.030316333103910773,
128      0.03234284386124316
129    ]
130  }
131}