Averaged Contractive State-Space Network / bench_report.json

Failed on benchmark

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": 0.01
 10    },
 11    "sweep": [
 12      {
 13        "cfg": {
 14          "lr": 0.001
 15        },
 16        "mean": 0.2699683979153633
 17      },
 18      {
 19        "cfg": {
 20          "lr": 0.003
 21        },
 22        "mean": 0.01333702343981713
 23      },
 24      {
 25        "cfg": {
 26          "lr": 0.01
 27        },
 28        "mean": 0.009317408432252705
 29      }
 30    ],
 31    "full": {
 32      "mean": 0.009518424805719405,
 33      "std": 0.004154286071903827,
 34      "per_seed": [
 35        0.013998116366565228,
 36        0.005631609819829464,
 37        0.013994917273521423,
 38        0.0036449902690947056,
 39        0.007559004705399275,
 40        0.006567755248397589,
 41        0.00925444532185793,
 42        0.01549655944108963
 43      ],
 44      "n": 8
 45    }
 46  },
 47  "idea": {
 48    "mean": 0.2998440582305193,
 49    "std": 0.07520865063647589,
 50    "per_seed": [
 51      0.39651888608932495,
 52      0.2933981418609619,
 53      0.41620975732803345,
 54      0.24436108767986298,
 55      0.27874669432640076,
 56      0.17241966724395752,
 57      0.3343149721622467,
 58      0.2627832591533661
 59    ],
 60    "n": 8
 61  },
 62  "comparison": {
 63    "delta_mean": 0.2903256334247999,
 64    "idea_wins": 0,
 65    "n_pairs": 8,
 66    "per_seed_diffs": [
 67      0.3825207697227597,
 68      0.28776653204113245,
 69      0.402214840054512,
 70      0.24071609741076827,
 71      0.2711876896210015,
 72      0.16585191199555993,
 73      0.3250605268403888,
 74      0.24728669971227646
 75    ],
 76    "p_value": 0.0081,
 77    "mde": 0.0650801969908384,
 78    "mde_rel_pct": 683.7286454344126,
 79    "verdict": "idea worse (significant)",
 80    "system_worked": false
 81  },
 82  "mechanism_signature": {
 83    "prediction": "trained averaged/fast state error decreases with eps; mu2 is negative",
 84    "rows": [
 85      {
 86        "eps": 0.5,
 87        "rms_state_error": 0.0011420742375776172
 88      },
 89      {
 90        "eps": 0.25,
 91        "rms_state_error": 0.0007367426878772676
 92      },
 93      {
 94        "eps": 0.125,
 95        "rms_state_error": 0.0004585969727486372
 96      },
 97      {
 98        "eps": 0.0625,
 99        "rms_state_error": 0.0007371138781309128
100      }
101    ],
102    "observed_loglog_slope": 0.2579025020059939,
103    "predicted_error_slope": 1.0,
104    "predicted_mu2_upper_bound": -1.35,
105    "observed_max_mu2": -1.5680792331695557,
106    "confirmed": false
107  },
108  "idea_sweep": [
109    {
110      "cfg": {
111        "lr": 0.001
112      },
113      "result": {
114        "mean": 0.7121945843100548,
115        "std": 0.0857891837384087,
116        "per_seed": [
117          0.8319963812828064,
118          0.690710723400116,
119          0.7957785129547119,
120          0.7097331285476685,
121          0.6804750561714172,
122          0.5556079149246216,
123          0.7924796342849731,
124          0.6407753229141235
125        ],
126        "n": 8
127      }
128    },
129    {
130      "cfg": {
131        "lr": 0.003
132      },
133      "result": {
134        "mean": 0.6083195172250271,
135        "std": 0.0816738853448453,
136        "per_seed": [
137          0.7198447585105896,
138          0.5920082926750183,
139          0.7039626836776733,
140          0.5942644476890564,
141          0.5862480998039246,
142          0.4565565288066864,
143          0.6692549586296082,
144          0.5444163680076599
145        ],
146        "n": 8
147      }
148    },
149    {
150      "cfg": {
151        "lr": 0.01
152      },
153      "result": {
154        "mean": 0.2998440582305193,
155        "std": 0.07520865063647589,
156        "per_seed": [
157          0.39651888608932495,
158          0.2933981418609619,
159          0.41620975732803345,
160          0.24436108767986298,
161          0.27874669432640076,
162          0.17241966724395752,
163          0.3343149721622467,
164          0.2627832591533661
165        ],
166        "n": 8
167      }
168    }
169  ],
170  "runtime_seconds": 79.51734733581543
171}