From 752dfb833b06a6fb974df892de560caf328ed1dd Mon Sep 17 00:00:00 2001 From: YurenHao0426 Date: Wed, 8 Apr 2026 18:03:16 -0500 Subject: paper v2.31: matched 30-epoch BP/DFA controls (was unsourced 0.609/0.308) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The §5 ¶3 BP-no-penalty value of 0.609 ± 0.004 and DFA-no-penalty value of 0.308 ± 0.014 turned out to be unsourced — they were carried over from a hardcoded comment in experiments/bp_with_penalty_control.py ("BP-trainable (3-seed mean): 0.609") that nobody had actually measured with a matched 30-epoch run. Ran the missing matched controls under the same recipe as BP+pen (lam=0, 30 epochs, AdamW 1e-3, wd 0.01, cosine schedule, batch 128, 3 seeds 42/123/456): BP no-pen 30ep: per-seed 0.5851, 0.5845, 0.5863 → 0.585 ± 0.001 (paper said 0.609 ± 0.004, off by 0.024) DFA no-pen 30ep: per-seed 0.3070, 0.2985, 0.2966 → 0.301 ± 0.005 (paper said 0.308 ± 0.014) Also re-grounded DFA+penalty 30ep using the dfa_pen_short 3-seed run (0.3593, 0.3610, 0.3604 → 0.360 ± 0.001), which is what the deep-cosine +0.155 figure was computed on. The paper had 0.363 ± 0.001 — that came from the 100-epoch run, not the 30-epoch run, so it was an apples-to- oranges comparison with BP+pen 30-ep. Paper changes (§5 ¶3): BP penalty cost: -8 pp → -5.5 pp DFA pen rescue: +5.5 → +5.9 pp DFA+pen margin vs frozen: +1.4 → +1.1 pp BP-to-DFA gap: 17 → 17.0 pp (unchanged) BP-to-SB gap: 7.7 → 7.7 pp (unchanged) BP-to-DFA gap is still the lower-bound credit-quality cost claim; 17 pp gap is unchanged in magnitude. Also updated: - §5 ¶1 prose: 0.363 → 0.360, 0.308 → 0.301 - §4 ¶4 prose: DFA+pen 0.363 → 0.360 - Appendix J Table 9 caption: 0.363 → 0.360, +9.0 → +9.3 pp gap to SB - Appendix L paragraph: +5.5 → +5.9 pp DFA penalty rescue - Figure 3 panel C bar values + title pen-cost annotation - New results/matched_30ep_control_summary.json as auditable record Page layout preserved: 9 main pages + refs p10, 18 total, 0 overfull. Co-Authored-By: Claude Opus 4.6 (1M context) --- results/dfa_nopen_30ep_3seed.log | 37 +++++++++++++++++++++++++++++++++++++ 1 file changed, 37 insertions(+) create mode 100644 results/dfa_nopen_30ep_3seed.log (limited to 'results/dfa_nopen_30ep_3seed.log') diff --git a/results/dfa_nopen_30ep_3seed.log b/results/dfa_nopen_30ep_3seed.log new file mode 100644 index 0000000..ec99d88 --- /dev/null +++ b/results/dfa_nopen_30ep_3seed.log @@ -0,0 +1,37 @@ +Using device: cuda:0 + +============================================================ +Seed 42 +============================================================ + +--- DFA --- + [DFA] Epoch 1: loss=2.0412, train=0.2521, test=0.2899 + [DFA] Epoch 10: loss=2.0028, train=0.2715, test=0.2862 + [DFA] Epoch 20: loss=1.9912, train=0.2790, test=0.3047 + [DFA] Epoch 30: loss=1.9860, train=0.2825, test=0.3070 + Final test acc: 0.3070 + +============================================================ +Seed 123 +============================================================ + +--- DFA --- + [DFA] Epoch 1: loss=2.0435, train=0.2482, test=0.2683 + [DFA] Epoch 10: loss=1.9956, train=0.2738, test=0.2943 + [DFA] Epoch 20: loss=1.9844, train=0.2785, test=0.2986 + [DFA] Epoch 30: loss=1.9814, train=0.2828, test=0.2985 + Final test acc: 0.2985 + +============================================================ +Seed 456 +============================================================ + +--- DFA --- + [DFA] Epoch 1: loss=2.0539, train=0.2465, test=0.2591 + [DFA] Epoch 10: loss=2.0115, train=0.2637, test=0.2887 + [DFA] Epoch 20: loss=1.9990, train=0.2728, test=0.2941 + [DFA] Epoch 30: loss=1.9955, train=0.2767, test=0.2966 + Final test acc: 0.2966 + +All results saved to results/dfa_no_penalty_30ep/results_cifar10.json +=== DONE === -- cgit v1.2.3