AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Fixup Initialization: Residual Learning Without Normalization

Hongyi Zhang, Yann N. Dauphin, Tengyu Ma

ICLR 2019 · Architectures and efficiency › RNNs and normalization · approach: Normalization layer theory · 102 citations (OpenAlex)

This paper tried a common change: Add mixup or CutMix

Add Mixup regularization to Fixup · ImageNet · test error (%): 27.6 -> 24.0 Improved

What 547 accepted papers report for this change

See this change in detailMatched by the wording of each paper's own result, read by AI from the full text. Papers report gains more often than failures, so this shows what gets published, not the chance a change works.

Read the paper

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
Replace BatchNorm with Fixup initializationResNet-110 · CIFAR-10test error (%): 6.61 -> 7.24Hurt
Replace Xavier initialization with Fixup initializationResNet-110 · CIFAR-10test error (%): 7.78 -> 7.24Improved
Add Mixup regularization to FixupResNet-50 · ImageNettest error (%): 27.6 -> 24.0Improved
Replace BatchNorm with Fixup and add MixupResNet-50 · ImageNettest error (%): 23.3 -> 24.0Hurt
Replace BatchNorm with Fixup and add MixupResNet-101 · ImageNettest error (%): 20.8 -> 21.4Hurt
Replace LayerNorm with FixupTransformer · IWSLT DE-ENBLEU: 34.2 -> 34.5Improved
Replace LayerNorm with FixupTransformer · WMT EN-DEBLEU: 29.3 -> 29.3No clear effect
Replace BatchNorm with Fixup and add Mixup + CutoutWideResNet-40-10 · CIFAR-10test error (%): 2.5 -> 2.3Improved
Replace BatchNorm with Fixup and add Mixup + CutoutWideResNet-16-12 · SVHNtest error (%): 1.4 -> 1.4No clear effect
Remove bias parameters (ablation)Fixup ResNet-110 · CIFAR-10training accuracyHurt
Use 0.1x the suggested initialization scale and no bias parameters (ablation)Fixup ResNet-110 · CIFAR-10training accuracyHurt
Use 10x the suggested initialization scale and no bias parameters (ablation)Fixup ResNet-110 · CIFAR-10training accuracyHurt

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