Moment-Sharp Spectral-Norm Control / results.json

Mechanism confirmed, baseline not beaten

Raw ⬇ ZIP
 1{
 2  "sanity": {
 3    "random_bound_minus_true_max": [
 4      [
 5        3,
 6        0.11826261871164886,
 7        1.2458506632298025
 8      ],
 9      [
10        5,
11        0.21836460116578976,
12        1.5893776858982678
13      ],
14      [
15        10,
16        0.47433732743115087,
17        3.6211179693465656
18      ],
19      [
20        24,
21        0.9592209839575727,
22        5.3156131681345435
23      ]
24    ],
25    "max_violation": -0.11826261871164886,
26    "clustered_exact_bound": 9.0,
27    "hutchinson_2000_probe_abs_error": [
28      0.0,
29      0.0
30    ]
31  },
32  "experiment": {
33    "weight_decay_baseline": {
34      "final_loss": 0.04228072240948677,
35      "mean_last20_loss": 0.04133407985791564,
36      "val_accuracy": 0.9666666666666667,
37      "max_grad_norm": 0.8295876979827881,
38      "seconds": 0.656698373000836,
39      "true_sigma_max": 3.12701061417523,
40      "moment_bound_sigma_max": 4.039054325668275,
41      "per_layer_true_sigma": [
42        3.12701061417523,
43        2.290517299144598,
44        1.491694104174947
45      ],
46      "per_layer_bound_sigma": [
47        4.039054325668275,
48        3.220958380344012,
49        2.073891055843045
50      ]
51    },
52    "K2_moment_rescale": {
53      "final_loss": 0.24251629412174225,
54      "mean_last20_loss": 0.2790691241621971,
55      "val_accuracy": 0.94,
56      "max_grad_norm": 0.778740644454956,
57      "seconds": 0.8602603210019879,
58      "true_sigma_max": 1.5439435366500303,
59      "moment_bound_sigma_max": 2.000000056124408,
60      "per_layer_true_sigma": [
61        1.5439435366500303,
62        1.4696971193323651,
63        1.487428996851836
64      ],
65      "per_layer_bound_sigma": [
66        2.000000056124408,
67        1.9999999828686161,
68        2.000000007668706
69      ]
70    }
71  }
72}