diff options
| -rw-r--r-- | docs/CLAIM_LARGEST_NOBP.md | 15 |
1 files changed, 15 insertions, 0 deletions
diff --git a/docs/CLAIM_LARGEST_NOBP.md b/docs/CLAIM_LARGEST_NOBP.md index ace8678..8bb69e5 100644 --- a/docs/CLAIM_LARGEST_NOBP.md +++ b/docs/CLAIM_LARGEST_NOBP.md @@ -53,3 +53,18 @@ gradient exactness (cos ~= 1 vs autograd reference) ยท analog-compatible trained - https://www.science.org/doi/10.1126/science.adi8474 - https://github.com/UbiquitousLearning/Backpropagation_Free_Training_Survey - https://arxiv.org/abs/2511.01061 (forward-only rigorous evaluation, 2025) + +## UPDATE 2026-07-11: Kerjan-Hoier-Scellier (Rain AI), arXiv:2606.03584 โ the closest relative, and a re-scoping. +"Training a PCN on ImageNet using EP" (June 2026): VGG10 PCN, full ImageNet, centered EP, +top-1 33.81 vs BP 32.35 (top-5 13.23 vs 12.20) โ **a real 1.0-1.5pp gap remains, at the hands of +EP's inventor**. Their construction IS the layered-energy bridge (their Eq. 10 == our cascade energy; +free phase == forward pass; single-equilibrium simplification == our single-sided legality; their +mod-PGD/asynchronous traversal ~= our fb solver). Vision classification only; hardware explicitly +punted ("remains unclear whether F_PCN could be efficiently realized in hardware"). Parameter count +not stated (VGG10-class, plausibly 20-130M) => **"largest NN today" is CONTESTED โ do not use.** +RE-SCOPED claims that hold today: largest/first TRANSFORMER fully trained without BP; first +AUTOREGRESSIVE LANGUAGE MODEL; the only zero-gap result (matched-tuning indistinguishability; their +best retains 1.0-1.5pp). At 1B+ the unqualified largest-NN claim reactivates regardless of their +param count. STRATEGIC: same lab as CET, funded by Rain AI + ARIA Scaling Compute; their refs +(mu-PC 100+ layers, Goemaere signal-decay, Qi step-size schedules; Momeni et al. Nature 2025 PNN +review) are our depth-scaling reading list and the venue precedent. |