AI Paper Atlas

Research map / Optimization / Implicit bias and generalization

Weight Conditioning for Smooth Optimization of Neural Networks

Hemanth Saratchandran, Thomas X Wang, Simon Lucey

ECCV 2024 · Optimization › Implicit bias and generalization · approach: Explicit and implicit regularization interplay · 2 citations (OpenAlex)

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
add equilibrated weight conditioning to batch normInception · CIFAR100Top-1 accuracy (%): 55.9 -> 58.7Improved
add equilibrated weight conditioning to batch normInception · CIFAR10Top-1 accuracy (%): 83.1 -> 83.3No clear effect
add equilibrated weight conditioning to batch normDenseNet · CIFAR10Top-1 accuracy (%): 70.5 -> 79.96Improved
add equilibrated weight conditioning to batch normDenseNet · CIFAR100Top-1 accuracy (%): 64.6 -> 72.4Improved
use equilibrated weight conditioning instead of weight standardizationDenseNet + BN · CIFAR10Top-1 accuracy (%): 75.9 -> 79.96Improved
add equilibrated weight conditioning to layernormViT-B · ImageNet1kTop-1 accuracy (%): 79.9 -> 80.2No clear effect
apply equilibrated weight conditioning to NeRF MLPNeRF · LLFF FernPSNR (dB): 27.51 -> 28.53Improved
apply equilibrated weight conditioning to NeRF MLPNeRF · LLFF FortressPSNR (dB): 33.16 -> 33.14No clear effect
apply equilibrated weight conditioning to NeRF MLPNeRF · LLFF (8 scenes)PSNR (dB): 29.6 -> 30. 20Improved

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