AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Mode Normalization

Lucas Deecke, Iain Murray, Hakan Bilen

ICLR 2019 · Architectures and efficiency › RNNs and normalization · approach: Batch normalization improvements · 8 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
replace batch normalization with mode normalizationLeNet · MNIST, CIFAR-10, SVHN, Fashion-MNIST multi-task; N=128test error (%): 26.91 ± 1.08 -> 23.16 ± 1.23Improved
use mode group normalization instead of group normalizationLeNet · MNIST, CIFAR-10, SVHN, Fashion-MNIST multi-task; N=4, 8, and 16test error (%): 32.15 ± 1.10 -> 31.30 ± 1.65; 28.60 ± 1.45 -> 26.83 ± 1.34; 27.63 ± 0.45 -> 26.00 ± 1.68Improved
replace batch normalization with mode normalizationNIN · CIFAR-10test error (%): 8.82 -> 8.42Improved
replace batch normalization with mode normalizationNIN · CIFAR-100test error (%): 32.30 -> 31.66Improved
replace batch normalization with mode normalizationVGG13 · CIFAR-10test error (%): 8.28 -> 7.79Improved
replace batch normalization with mode normalizationVGG13 · CIFAR-100test error (%): 31.15 -> 30.06Improved
replace batch normalization with mode normalizationResNet20 · CIFAR-10test error (%): 8.44 -> 7.99Improved
replace batch normalization with mode normalizationResNet20 · CIFAR-100test error (%): 31.56 -> 30.53Improved
replace batch normalization with mode normalizationResNet18 · ILSVRC12top-1 error (%): 30.25 -> 30.07Improved
replace batch normalization with mode normalizationResNet18 · ILSVRC12top-5 error (%): 10.90 -> 10.65Improved
increase the number of modes from K=2 to K=4 (ablation)LeNet with mode normalization · MNIST, CIFAR-10, SVHN, Fashion-MNIST multi-task; N=128test error (%): 23.16 ± 1.23 -> 24.25 ± 0.71Hurt
increase the number of modes from K=2 to K=6 (ablation)LeNet with mode normalization · MNIST, CIFAR-10, SVHN, Fashion-MNIST multi-task; N=128test error (%): 23.16 ± 1.23 -> 25.12 ± 1.48Hurt

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