AI Paper Atlas

Research map / Architectures and efficiency / CNN design

Impact of Aliasing on Generalization in Deep Convolutional Networks

Cristina Vasconcelos, Hugo Larochelle, Vincent Dumoulin, Rob Romijnders, Nicolas Le Roux, Ross Goroshin

ICCV 2021 · Architectures and efficiency › CNN design · approach: Theory of convolutional inductive bias · 33 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
add pre-filter to strided skip connections (ablation)ResNet-50 · ImageNet val (90 epochs)top-1 accuracy (%): 76.49 -> 77.14Improved
add post-filter to strided skip connections (ablation)ResNet-50 · ImageNet val (90 epochs)top-1 accuracy (%): 76.49 -> 77.15Improved
insert low-pass filters in all convolutional layers (ablation)ResNet-50 · ImageNet val (90 epochs)top-1 accuracy (%): 76.49 -> 61.00Hurt
anti-alias the conv1 layer that already has a 7x7 kernel (ablation)ResNet-50 · ImageNet val (90 epochs)top-1 accuracy (%)Hurt
add post-filter anti-aliasing at all strided critical paths (ablation)ResNet-50 · ImageNet val (90 epochs)top-1 accuracy (%): 76.49 -> 77.47Improved
add anti-aliasing (ablation)ResNet-50 · ImageNet val (180 epochs)top-1 accuracy (%): 77.36 -> 77.76Improved
add random augmentation (ablation)ResNet-50 · ImageNet val (180 epochs)top-1 accuracy (%): 77.36 -> 77.32No clear effect
add anti-aliasing on top of random augmentation (ablation)ResNet-50 · ImageNet val (180 epochs)top-1 accuracy (%): 77.32 -> 78.45Improved
add anti-aliasing (k=3) (ablation)EfficientNet-B0 · ImageNet val (3-seed mean, appendix table)top-1 accuracy (%): 76.40 -> 76.65Improved
add anti-aliasing (k=3) with random augmentation (ablation)EfficientNet-B0 · ImageNet val (3-seed mean, appendix table)top-1 accuracy (%): 76.98 -> 77.17Improved
use anti-aliasing instead of Zhang's filtersResNet-50 · ImageNet-Rtop-1 accuracy (%): 24.4 -> 25.2Improved
add anti-aliasing and GELU to SURSUR · Meta-Dataset (all training sources)average rank: 3.40 -> 1.65Improved

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