Resonance-Aware Stochastic RNN Control / bench_report.json

✓✓ Beats tuned baseline

Raw ⬇ ZIP
  1{
  2  "bench_version": 1,
  3  "track": "dynamics",
  4  "model": "rnn_small",
  5  "metric_direction": "lower is better",
  6  "n_seeds": 8,
  7  "baseline": {
  8    "best_cfg": {
  9      "lr": 0.006,
 10      "weight_decay": 0.0001
 11    },
 12    "sweep": [
 13      {
 14        "cfg": {
 15          "lr": 0.001,
 16          "weight_decay": 0.0
 17        },
 18        "mean": 0.004298779065720737
 19      },
 20      {
 21        "cfg": {
 22          "lr": 0.001,
 23          "weight_decay": 0.0001
 24        },
 25        "mean": 0.0042995543335564435
 26      },
 27      {
 28        "cfg": {
 29          "lr": 0.003,
 30          "weight_decay": 0.0
 31        },
 32        "mean": 0.0034615940821822733
 33      },
 34      {
 35        "cfg": {
 36          "lr": 0.003,
 37          "weight_decay": 0.0001
 38        },
 39        "mean": 0.003432174737099558
 40      },
 41      {
 42        "cfg": {
 43          "lr": 0.006,
 44          "weight_decay": 0.0
 45        },
 46        "mean": 0.0018282315868418664
 47      },
 48      {
 49        "cfg": {
 50          "lr": 0.006,
 51          "weight_decay": 0.0001
 52        },
 53        "mean": 0.0017760894843377173
 54      }
 55    ],
 56    "full": {
 57      "mean": 0.0020704464986920357,
 58      "std": 0.00039353273372166414,
 59      "per_seed": [
 60        0.0019586661364883184,
 61        0.0013749259524047375,
 62        0.00156879099085927,
 63        0.002201974857598543,
 64        0.0023778362665325403,
 65        0.002474326640367508,
 66        0.0020717033185064793,
 67        0.0025353478267788887
 68      ],
 69      "n": 8
 70    }
 71  },
 72  "idea": {
 73    "mean": 0.0014319315523607656,
 74    "std": 0.00040801966999132736,
 75    "per_seed": [
 76      0.00139222526922822,
 77      0.0008285212097689509,
 78      0.0013016352895647287,
 79      0.001789333182387054,
 80      0.000968311564065516,
 81      0.0018908914644271135,
 82      0.0012502360623329878,
 83      0.002034298377111554
 84    ],
 85    "n": 8
 86  },
 87  "comparison": {
 88    "delta_mean": -0.00063851494633127,
 89    "idea_wins": 8,
 90    "n_pairs": 8,
 91    "per_seed_diffs": [
 92      -0.0005664408672600985,
 93      -0.0005464047426357865,
 94      -0.00026715570129454136,
 95      -0.0004126416752114892,
 96      -0.0014095247024670243,
 97      -0.0005834351759403944,
 98      -0.0008214672561734915,
 99      -0.0005010494496673346
100    ],
101    "p_value": 0.0081,
102    "mde": 0.0002917931104320964,
103    "mde_rel_pct": 14.093245617137706,
104    "verdict": "idea better (significant)",
105    "system_worked": true
106  },
107  "mechanism_signature": {
108    "predicted_vs_observed": {
109      "target_radius": 0.9,
110      "observed_radius_mean": 1.5063634291291237,
111      "fitted_mode_rows": [
112        {
113          "epoch": 0,
114          "fitted_radius": 0.8541139240506328,
115          "fitted_omega": 0.05
116        },
117        {
118          "epoch": 11,
119          "fitted_radius": 0.894367088607595,
120          "fitted_omega": 0.05
121        }
122      ]
123    },
124    "confirmed": false
125  },
126  "idea_trials": [
127    {
128      "cfg": {
129        "lr": 0.006,
130        "weight_decay": 0.0001,
131        "r_max": 0.9,
132        "beta": 0.02
133      },
134      "result": {
135        "mean": 0.0014319315523607656,
136        "std": 0.00040801966999132736,
137        "per_seed": [
138          0.00139222526922822,
139          0.0008285212097689509,
140          0.0013016352895647287,
141          0.001789333182387054,
142          0.000968311564065516,
143          0.0018908914644271135,
144          0.0012502360623329878,
145          0.002034298377111554
146        ],
147        "n": 8
148      }
149    },
150    {
151      "cfg": {
152        "lr": 0.003,
153        "weight_decay": 0.0001,
154        "r_max": 0.9,
155        "beta": 0.02
156      },
157      "result": {
158        "mean": 0.0028593493334483355,
159        "std": 0.001481906654941589,
160        "per_seed": [
161          0.0025175209157168865,
162          0.0032294574193656445,
163          0.0012714689364656806,
164          0.006308244541287422,
165          0.002197133842855692,
166          0.003583323210477829,
167          0.0020109813194721937,
168          0.0017566644819453359
169        ],
170        "n": 8
171      }
172    },
173    {
174      "cfg": {
175        "lr": 0.006,
176        "weight_decay": 0.0001,
177        "r_max": 0.9,
178        "beta": 0.02
179      },
180      "result": {
181        "mean": 0.0014319315523607656,
182        "std": 0.00040801966999132736,
183        "per_seed": [
184          0.00139222526922822,
185          0.0008285212097689509,
186          0.0013016352895647287,
187          0.001789333182387054,
188          0.000968311564065516,
189          0.0018908914644271135,
190          0.0012502360623329878,
191          0.002034298377111554
192        ],
193        "n": 8
194      }
195    }
196  ],
197  "math_sanity": {
198    "true_radius": 0.93,
199    "fitted_radius": 0.9346202531645569,
200    "true_omega": 0.42,
201    "fitted_omega": 0.4275280898876404,
202    "radius_abs_error": 0.004620253164556831,
203    "frequency_abs_error": 0.007528089887640421,
204    "pass": true
205  }
206}