AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

GradInit: Learning to Initialize Neural Networks for Stable and Efficient Training

Chen Zhu, Renkun Ni, Zheng Xu, Kezhi Kong, W Ronny Huang, Tom Goldstein

NeurIPS 2021 · Architectures and efficiency › RNNs and normalization · approach: Layer-wise analysis of language models · 9 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
use GradInit initialization instead of KaimingVGG-19 w/ BN · CIFAR-10best test accuracy (%): 94.4 -> 95.1Improved
use GradInit instead of Kaiming for 1202-layer ResNetResNet-1202 w/ BN · CIFAR-10best test accuracy (%): 94.4 -> 96.2Improved
use GradInit instead of Kaiming for ResNet-50 on ImageNetResNet-50 w/ BN · ImageNettop-1 accuracy (%): 75.9 -> 76.2Improved
use GradInit without warmup for Post-LN Transformer trained with SGD vs AdminPost-LN Transformer · IWSLT-14 De-EnBLEU: 33.7 -> 35.6Improved
also rescale BN scale parameters (vs fixing BN) (ablation)GradInit · CIFAR-10best test accuracy (%): 94.6 -> 95.1 (VGG-19 BN); 94.7 -> 95.4 (ResNet-110 BN)Improved
also rescale linear layers (vs only BN) (ablation)GradInit · CIFAR-10best test accuracy (%): 94.4 -> 95.1 (VGG-19 BN)Improved
use GradInit's one-step-loss objective instead of simply learning scalars (ablation)GradInit · CIFAR-10best test accuracy (%): 10.0 -> 95.1 (VGG-19 BN); 94.7 -> 95.4 (ResNet-110 BN)Improved
use gradient-norm constraint (gamma=1) instead of gradient penalty lambda=1 (ablation)GradInit · CIFAR-10best test accuracy (%): 74.7 -> 94.7 (VGG-19 w/o BN)Improved
use 50% overlap between S and S~ instead of independent batches (ablation)GradInit · CIFAR-10first-epoch accuracy (%): 21.9 -> 29.3Improved
use 50% overlap instead of identical batches (S~ = S) (ablation)GradInit · CIFAR-10best test accuracy (%): 94.5 -> 94.7Improved
keep gradient norm constraint when S~ = S (ablation)GradInit · CIFAR-10best test accuracy (%): 93.70 -> 94.78Improved
add skip scaling w_skip to GradInit Transformer (ablation)GradInit with Adam · IWSLT-14 De-EnBLEU: 36.1 -> 36.0Mixed

Later papers that compared with this method

1 later accepted paper reports Gradinit as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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