First-Hit Interacting Optimizer / bench_report.json

Failed on benchmark

Raw ⬇ ZIP
  1{
  2  "bench_version": 1,
  3  "track": "tabular",
  4  "model": "mlp_tiny",
  5  "metric_direction": "lower is better",
  6  "n_seeds": 8,
  7  "baseline": {
  8    "best_cfg": {
  9      "lr": 0.003,
 10      "weight_decay": 0.0
 11    },
 12    "sweep": [
 13      {
 14        "cfg": {
 15          "lr": 0.001,
 16          "weight_decay": 0.0
 17        },
 18        "mean": 55.30739402770996
 19      },
 20      {
 21        "cfg": {
 22          "lr": 0.001,
 23          "weight_decay": 0.0001
 24        },
 25        "mean": 55.30739402770996
 26      },
 27      {
 28        "cfg": {
 29          "lr": 0.003,
 30          "weight_decay": 0.0
 31        },
 32        "mean": 9.189739227294922
 33      },
 34      {
 35        "cfg": {
 36          "lr": 0.003,
 37          "weight_decay": 0.0001
 38        },
 39        "mean": 9.189739227294922
 40      },
 41      {
 42        "cfg": {
 43          "lr": 0.01,
 44          "weight_decay": 0.0
 45        },
 46        "mean": 15.7157564163208
 47      },
 48      {
 49        "cfg": {
 50          "lr": 0.01,
 51          "weight_decay": 0.0001
 52        },
 53        "mean": 15.7157564163208
 54      }
 55    ],
 56    "full": {
 57      "mean": 9.629322290420532,
 58      "std": 1.0272830694739892,
 59      "per_seed": [
 60        8.371176719665527,
 61        9.78935432434082,
 62        10.369681358337402,
 63        8.228744506835938,
 64        8.8460054397583,
 65        11.477396011352539,
 66        10.141501426696777,
 67        9.810718536376953
 68      ],
 69      "n": 8
 70    }
 71  },
 72  "idea": {
 73    "mean": 9.629322290420532,
 74    "std": 1.0272830694739892,
 75    "per_seed": [
 76      8.371176719665527,
 77      9.78935432434082,
 78      10.369681358337402,
 79      8.228744506835938,
 80      8.8460054397583,
 81      11.477396011352539,
 82      10.141501426696777,
 83      9.810718536376953
 84    ],
 85    "n": 8
 86  },
 87  "comparison": {
 88    "delta_mean": 0.0,
 89    "idea_wins": 0,
 90    "n_pairs": 8,
 91    "per_seed_diffs": [
 92      0.0,
 93      0.0,
 94      0.0,
 95      0.0,
 96      0.0,
 97      0.0,
 98      0.0,
 99      0.0
100    ],
101    "p_value": 1.0,
102    "mde": 0.0,
103    "mde_rel_pct": 0.0,
104    "verdict": "no measurable effect",
105    "system_worked": false
106  },
107  "mechanism_signature": {
108    "prediction": "unnormalized coherent interaction",
109    "best_cfg": {
110      "lr": 0.003,
111      "alpha": 0.0,
112      "normalized": true
113    },
114    "mechanism_signature": {
115      "prediction": "unnormalized coherent force should yield algebraic first-hit acceleration (ideal slope -1), unlike normalized bounded force",
116      "observed_NN_scale": [
117        {
118          "N": 2,
119          "mean_test_metric": 9.453511953353882,
120          "mean_first_hit_epoch": 2.25,
121          "mean_force_to_gradient": 0.0
122        },
123        {
124          "N": 4,
125          "mean_test_metric": 9.189739227294922,
126          "mean_first_hit_epoch": 2.25,
127          "mean_force_to_gradient": 0.0
128        },
129        {
130          "N": 8,
131          "mean_test_metric": 9.189739227294922,
132          "mean_first_hit_epoch": 2.0,
133          "mean_force_to_gradient": 0.0
134        }
135      ],
136      "observed_loglog_first_hit_slope": -0.08496250072115638,
137      "confirmed": false
138    }
139  },
140  "idea_sweep": [
141    {
142      "cfg": {
143        "lr": 0.001,
144        "alpha": 0.0,
145        "normalized": true
146      },
147      "mean": 54.075210094451904,
148      "per_seed": [
149        75.81849670410156,
150        49.25731658935547,
151        58.83588790893555,
152        37.317874908447266,
153        41.6431770324707,
154        45.07895278930664,
155        62.11760711669922,
156        62.53236770629883
157      ]
158    },
159    {
160      "cfg": {
161        "lr": 0.001,
162        "alpha": 0.1,
163        "normalized": true
164      },
165      "mean": 50.048232078552246,
166      "per_seed": [
167        74.95216369628906,
168        43.78724670410156,
169        53.71392822265625,
170        33.512508392333984,
171        37.85257339477539,
172        41.05773162841797,
173        56.02418899536133,
174        59.48551559448242
175      ]
176    },
177    {
178      "cfg": {
179        "lr": 0.001,
180        "alpha": 0.1,
181        "normalized": false
182      },
183      "mean": 50.5640082359314,
184      "per_seed": [
185        75.70262145996094,
186        43.93588638305664,
187        54.4977912902832,
188        33.584346771240234,
189        38.557464599609375,
190        41.220726013183594,
191        56.3896369934082,
192        60.623592376708984
193      ]
194    },
195    {
196      "cfg": {
197        "lr": 0.003,
198        "alpha": 0.0,
199        "normalized": true
200      },
201      "mean": 9.629322290420532,
202      "per_seed": [
203        8.371176719665527,
204        9.78935432434082,
205        10.369681358337402,
206        8.228744506835938,
207        8.8460054397583,
208        11.477396011352539,
209        10.141501426696777,
210        9.810718536376953
211      ]
212    },
213    {
214      "cfg": {
215        "lr": 0.003,
216        "alpha": 0.1,
217        "normalized": true
218      },
219      "mean": 9.833800315856934,
220      "per_seed": [
221        8.321816444396973,
222        9.017751693725586,
223        10.796687126159668,
224        8.990129470825195,
225        9.358880043029785,
226        11.234187126159668,
227        10.641441345214844,
228        10.30950927734375
229      ]
230    },
231    {
232      "cfg": {
233        "lr": 0.003,
234        "alpha": 0.1,
235        "normalized": false
236      },
237      "mean": 9.8294016122818,
238      "per_seed": [
239        8.701092720031738,
240        9.271023750305176,
241        10.7781343460083,
242        9.142780303955078,
243        9.247992515563965,
244        11.352561950683594,
245        9.731724739074707,
246        10.409902572631836
247      ]
248    },
249    {
250      "cfg": {
251        "lr": 0.01,
252        "alpha": 0.0,
253        "normalized": true
254      },
255      "mean": 16.260348558425903,
256      "per_seed": [
257        13.586896896362305,
258        15.982109069824219,
259        19.061920166015625,
260        14.232099533081055,
261        14.604202270507812,
262        17.730175018310547,
263        16.55130958557129,
264        18.334075927734375
265      ]
266    },
267    {
268      "cfg": {
269        "lr": 0.01,
270        "alpha": 0.1,
271        "normalized": true
272      },
273      "mean": 17.912508606910706,
274      "per_seed": [
275        17.905563354492188,
276        15.030466079711914,
277        19.168237686157227,
278        15.649327278137207,
279        16.334321975708008,
280        20.19869613647461,
281        17.052860260009766,
282        21.960596084594727
283      ]
284    },
285    {
286      "cfg": {
287        "lr": 0.01,
288        "alpha": 0.1,
289        "normalized": false
290      },
291      "mean": 18.415680766105652,
292      "per_seed": [
293        17.90578842163086,
294        17.756229400634766,
295        20.10426902770996,
296        15.55553150177002,
297        16.786943435668945,
298        20.642580032348633,
299        16.461769104003906,
300        22.112335205078125
301      ]
302    }
303  ],
304  "equal_budget": {
305    "baseline": "4 independent Adam particles",
306    "idea": "4 interacting Adam particles",
307    "epochs_per_particle": 10,
308    "batch": 128
309  }
310}