From c09ae4244033a7a2703f0c36279d598ca869a95f Mon Sep 17 00:00:00 2001 From: YurenHao0426 Date: Tue, 24 Mar 2026 01:44:34 -0500 Subject: Add CIFAR deltaL test (failed) and pivot design memo MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit - CIFAR deltaL: s=grad_hL CE (dim=512) -> acc=17.2%, Gamma≈0 Confirms scalar value field has dimensionality bottleneck on CIFAR - Pivot memo: direct vector credit field a_phi(h,t,s) -> R^d Trained with perturbation-based target, avoids curvature problem Still satisfies no hidden BP anchor constraint Co-Authored-By: Claude Opus 4.6 (1M context) --- NOTE.md | 12 ++++++++++++ 1 file changed, 12 insertions(+) (limited to 'NOTE.md') diff --git a/NOTE.md b/NOTE.md index 8d6091d..774b72a 100644 --- a/NOTE.md +++ b/NOTE.md @@ -153,3 +153,15 @@ wr=0.5 -> worst Gamma (0.23) but best acc (0.66). Clear tradeoff between credit quality and accuracy. Best single config: deltaL + tgw=1.0 + wr=0.05 -> **Gamma=0.768, rho=0.691** + +### CIFAR deltaL Test +deltaL conditioning (s=grad_{h_L} CE, dim=512) on CIFAR L=4: FAILED. +Acc=17.2%, Gamma≈0, rho≈0. The 512-dim conditioning is too high-dimensional +for the value net. Confirms the scalar V approach has a dimensionality bottleneck. + +### Pivot Recommendation: Direct Vector Credit Field +See `report_explore/MEMO_pivot_vector_field.md`. +Instead of V_phi -> grad_h V, learn a_phi(h_l, t_l, s) -> R^d directly. +Train with perturbation-based target: match to actual loss change. +Still satisfies no hidden BP anchor constraint. +Minimal test: synthetic alpha=1.0, L=4 with M=4 perturbation directions. -- cgit v1.2.3