AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Revisiting Batch Norm Initialization

Jim Davis, Logan Frank

ECCV 2022 · Architectures and efficiency › RNNs and normalization · approach: Batch normalization improvements

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
initialize BN gamma at 0.1 with gamma lr divided by 100ResNet-18 · CIFAR-10, lr 0.1test accuracy (%): 89.44 -> 90.80Improved
initialize BN gamma at 0.05 with gamma lr divided by 100ResNet-18 · CIFAR-10, lr 0.01test accuracy (%): 84.64 -> 88.84Improved
initialize BN gamma at 0.01 with gamma lr divided by 100ResNet-18 · CIFAR-10, lr 0.001test accuracy (%): 71.32 -> 80.37Improved
add gamma below 1 initialization on top of gamma lr reduction (ablation)ResNet-18 · CIFAR-10, lr 0.1test accuracy (%): 89.81 -> 90.80Improved
add gamma lr reduction (c=100) to gamma<1 initialization (ablation)ResNet-18 · CIFAR-10test accuracy (%)Improved
increase learning rate for BN parameters only (ablation)BN parameters · CIFAR-10test accuracy (%)Hurt
scale only normalized data (standard form) instead of scaling beta too (A1, A2) (ablation)BN affine · CIFAR-10, low lrtest accuracy (%): up to 7% gainImproved
apply method on ResNet-101ResNet-101 · CIFAR-10, lr 0.1test accuracy (%): 88.28 -> 91.58Improved
apply method on ResNeXt-50ResNeXt-50 · CIFAR-10, lr 0.1test accuracy (%): 88.60 -> 92.04Improved
apply method to Stanford Cars at lr 0.01ResNet-50 · Stanford Cars, lr 0.01test accuracy (%): 51.86 -> 51.18No clear effect
use proposed gamma init instead of RBN scaleResNet-34 · CIFAR-100, lr 0.01test accuracy (%): 58.95 -> 64.01Improved
use instance-specific statistics in RBN/IEBN scaling (ablation)RBN/IEBN · CIFAR-10test accuracy (%): 84.72 -> 84.50 (RBN vs RBN-, lr 0.01)No clear effect

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