AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Weight Feedback Computes the Exact Jacobian Transpose in Modern Deep Networks

Junlong Shen, Xingyu Li

ECCV 2026 · Architectures and efficiency › RNNs and normalization · approach: Backprop-free credit assignment

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use W^T feedback instead of random matrix B (ablation)WF-Act-PC · VGG-5/CIFAR-10test accuracy (%): 27.42 -> 88.25Improved
add sigma' correction to W^T feedback (ablation)WF-Act-PC · VGG-5/CIFAR-10test accuracy (%): 81.62 -> 88.25Improved
replace local transport with autograd J^T (ablation)WF-Act-PC · VGG-5/CIFAR-10test accuracy (%): 88.25 -> 89.01No clear effect
add soft spectral norm clipping (ablation)WF-Act-PC · VGG-5/CIFAR-10test accuracy (%): 87.40 -> 88.25Improved
increase inner-loop iterations from T=1 to T=10 (ablation)WF-Act-PC · VGG-5/CIFAR-10test accuracy (%): 87.64 -> 88.29Improved
increase inner-loop iterations from T=20 to T=40 (ablation)WF-Act-PC · VGG-5/CIFAR-10test accuracy (%): 88.25 -> 72.01Hurt
add normalization gain s correction with BatchNorm (ablation)WF-Act-PC · BN-VGG-5/CIFAR-10test accuracy (%): 86.03 -> 87.60Improved
add sigma' correction with BatchNorm (ablation)WF-Act-PC · BN-VGG-5/CIFAR-10test accuracy (%): 84.48 -> 87.60Improved
use exact argmax unpooling instead of nearest-neighbour for MaxPool (ablation)WF-Act-PC · VGG-5/CIFAR-10test accuracy (%): 88.32 -> 89.58Improved
vary SNC threshold tau in {2,3,5,inf} (ablation)WF-Act-PC · VGG-5/CIFAR-10 (single seed)test accuracy (%): 87.9/88.3/87.9/87.4No clear effect
extend inner loop to T=30 (ablation)WF-Act-PC · VGG-5/CIFAR-10 (single seed)test accuracy (%): 88.2No clear effect
use tuned BP instead of WF-Act-PC (VGG-9) (ablation)BP (tuned) · CIFAR-100 VGG-9test accuracy (%): 71.06 -> 71.90Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy