AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Layer-wise Conditioning Analysis in Exploring the Learning Dynamics of DNNs

Lei Huang, Jie Qin, Li Liu, Fan Zhu, Ling Shao

ECCV 2020 · Architectures and efficiency › RNNs and normalization · approach: Representation geometry and collapse

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 one BN layer before the last linear layerResNet 1202-layer · CIFAR-10test error: 9.42 ± 3.10 -> 5.68 ± 0.14Improved
add one BN layer before the last linear layerResNet 56-layer · CIFAR-10test error: 7.52 ± 0.30 -> 6.50 ± 0.22Improved
add one BN layer before the last linear layerResNet 110-layer · CIFAR-10test error: 6.89 ± 0.52 -> 6.10 ± 0.09Improved
add one BN layer before the last linear layerResNet 1202-layer · CIFAR-100test error: 30.49 ± 4.44 -> 25.51 ± 0.27Improved
add one BN layer before the last linear layerResNet 56-layer · CIFAR-100test error: 29.60 ± 0.41 -> 28.82 ± 0.38Improved
add the extra BN layer before final pooling (ablation)PreResNet-style 1202-layer · CIFAR-10test error: 7.89 ± 0.58 -> 6.07 ± 0.10Improved
add the extra BN layer before final pooling (ablation)PreResNet-style 1202-layer · CIFAR-100test error: 30.07 ± 2.04 -> 26.23 ± 0.26Improved
add the extra BN layer before final pooling (ablation)PreResNet-style 56-layer · CIFAR-10test error: 6.75 ± 0.26 -> 6.89 ± 0.09No clear effect
add one BN layer before the last linear layerResNet 18-layer · ImageNet valtop-1 error: 29.78 -> 29.38Improved
add one BN layer before the last linear layerResNet 18-layer with WD=0.0002 · ImageNet valtop-1 error: 30.07 -> 28.96Improved
add one BN layer before the last linear layerResNet 50-layer · ImageNet valtop-1 error: 23.97 -> 23.76Improved
add one BN layer before the last linear layerVGG-style without BN, 56-layer · CIFAR-10test error: 89.93 ± 0.12 -> 23.40 ± 1.74Improved

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