Laplacian-Coherence Graph Minibatches / bench_report.json

✓✓ Beats tuned baseline

Raw ⬇ ZIP
  1{
  2  "bench_version": 1,
  3  "track": "laplacian_coherence_graph",
  4  "model": "local_gcn",
  5  "metric_direction": "lower is better",
  6  "n_seeds": 8,
  7  "baseline": {
  8    "best_cfg": {
  9      "lr": 0.006,
 10      "wd": 0.0,
 11      "epochs": 18,
 12      "batch": 32,
 13      "cand": 64
 14    },
 15    "sweep": [
 16      {
 17        "cfg": {
 18          "lr": 0.001,
 19          "wd": 0.0,
 20          "epochs": 18,
 21          "batch": 32,
 22          "cand": 64
 23        },
 24        "mean": 0.4312499985098839
 25      },
 26      {
 27        "cfg": {
 28          "lr": 0.001,
 29          "wd": 0.0001,
 30          "epochs": 18,
 31          "batch": 32,
 32          "cand": 64
 33        },
 34        "mean": 0.4312499985098839
 35      },
 36      {
 37        "cfg": {
 38          "lr": 0.003,
 39          "wd": 0.0,
 40          "epochs": 18,
 41          "batch": 32,
 42          "cand": 64
 43        },
 44        "mean": 0.34687499701976776
 45      },
 46      {
 47        "cfg": {
 48          "lr": 0.003,
 49          "wd": 0.0001,
 50          "epochs": 18,
 51          "batch": 32,
 52          "cand": 64
 53        },
 54        "mean": 0.3437499925494194
 55      },
 56      {
 57        "cfg": {
 58          "lr": 0.006,
 59          "wd": 0.0,
 60          "epochs": 18,
 61          "batch": 32,
 62          "cand": 64
 63        },
 64        "mean": 0.30937499925494194
 65      },
 66      {
 67        "cfg": {
 68          "lr": 0.006,
 69          "wd": 0.0001,
 70          "epochs": 18,
 71          "batch": 32,
 72          "cand": 64
 73        },
 74        "mean": 0.3124999962747097
 75      }
 76    ],
 77    "full": {
 78      "mean": 0.3468749988824129,
 79      "std": 0.09264978863573903,
 80      "per_seed": [
 81        0.23749999701976776,
 82        0.3499999940395355,
 83        0.3125,
 84        0.3375000059604645,
 85        0.5249999761581421,
 86        0.21250000596046448,
 87        0.4000000059604645,
 88        0.4000000059604645
 89      ],
 90      "n": 8
 91    }
 92  },
 93  "idea": {
 94    "mean": 0.29999999701976776,
 95    "std": 0.08028270303663049,
 96    "per_seed": [
 97      0.22499999403953552,
 98      0.36250001192092896,
 99      0.20000000298023224,
100      0.32499998807907104,
101      0.38749998807907104,
102      0.17499999701976776,
103      0.3499999940395355,
104      0.375
105    ],
106    "n": 8
107  },
108  "comparison": {
109    "delta_mean": -0.04687500186264515,
110    "idea_wins": 7,
111    "n_pairs": 8,
112    "per_seed_diffs": [
113      -0.012500002980232239,
114      0.012500017881393433,
115      -0.11249999701976776,
116      -0.012500017881393433,
117      -0.13749998807907104,
118      -0.037500008940696716,
119      -0.050000011920928955,
120      -0.025000005960464478
121    ],
122    "p_value": 0.0315,
123    "mde": 0.0435446955236248,
124    "mde_rel_pct": 12.553425776986021,
125    "verdict": "idea better (significant)",
126    "system_worked": true
127  },
128  "mechanism_signature": {
129    "custom_track": {
130      "name": "laplacian_coherence_graph",
131      "file": "graph_track.py",
132      "domain": "graph-nn"
133    },
134    "observed_best_cfg": {
135      "lr": 0.006,
136      "wd": 0.0,
137      "epochs": 18,
138      "batch": 32,
139      "cand": 64
140    },
141    "idea_sweep": [
142      {
143        "cfg": {
144          "lr": 0.001,
145          "wd": 0.0,
146          "epochs": 18,
147          "batch": 32,
148          "cand": 64
149        },
150        "result": {
151          "mean": 0.4546874985098839,
152          "std": 0.10642219395208924,
153          "per_seed": [
154            0.4375,
155            0.4625000059604645,
156            0.4749999940395355,
157            0.42500001192092896,
158            0.612500011920929,
159            0.25,
160            0.5874999761581421,
161            0.38749998807907104
162          ],
163          "n": 8
164        }
165      },
166      {
167        "cfg": {
168          "lr": 0.003,
169          "wd": 0.0,
170          "epochs": 18,
171          "batch": 32,
172          "cand": 64
173        },
174        "result": {
175          "mean": 0.3671875,
176          "std": 0.10037770716979075,
177          "per_seed": [
178            0.32499998807907104,
179            0.4000000059604645,
180            0.21250000596046448,
181            0.4000000059604645,
182            0.48750001192092896,
183            0.22499999403953552,
184            0.5,
185            0.38749998807907104
186          ],
187          "n": 8
188        }
189      },
190      {
191        "cfg": {
192          "lr": 0.006,
193          "wd": 0.0,
194          "epochs": 18,
195          "batch": 32,
196          "cand": 64
197        },
198        "result": {
199          "mean": 0.29999999701976776,
200          "std": 0.08028270303663049,
201          "per_seed": [
202            0.22499999403953552,
203            0.36250001192092896,
204            0.20000000298023224,
205            0.32499998807907104,
206            0.38749998807907104,
207            0.17499999701976776,
208            0.3499999940395355,
209            0.375
210          ],
211          "n": 8
212        }
213      }
214    ],
215    "mechanism_signature": {
216      "metric": 0.22499999403953552,
217      "observed_coherent_coverage": 2.0,
218      "observed_uniform_coverage": 2.0,
219      "predicted_candidate_pool_coverage": 2.0,
220      "confirmed": true
221    }
222  }
223}