AI Paper Atlas

Research map / Architectures and efficiency / CNN design

Truly Scale-Equivariant Deep Nets with Fourier Layers

Md Ashiqur Rahman, Raymond A. Yeh

NeurIPS 2023 · Architectures and efficiency › CNN design · approach: Group symmetry equivariant networks · 1 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 scale-consistency loss (ablation)Ours · MNIST-scale 5000 samplesaccuracy: 0.9831 -> 0.9835Improved
Add scale-consistency loss (ablation)Ours · MNIST-scale 5000 samplesscale-consistency: 0.9150 -> 0.9296Improved
Add scale-consistency loss (ablation)Ours · MNIST-scale 1000 samplesscale-consistency: 0.8144 -> 0.8183Improved
Add spatially local Fourier layers (ablation)Ours · MNIST-scale idealaccuracy: 0.9835 -> 0.9889Improved
Add scale-equivariant ReLU (ablation)Ours · MNIST-scale idealequivariance error: 7.32 -> 0.00Improved
Add scale-equivariant ReLU (ablation)Ours · MNIST-scale idealaccuracy: 0.9897 -> 0.9889No clear effect
Ideal upsampling instead of zero-padding for baselines (ablation)SI-CovNet · STL10-scale idealaccuracy: 0.5588 -> 0.6722Improved
Ideal upsampling instead of zero-padding for baselines (ablation)SS-CNN · STL10-scale idealaccuracy: 0.4788 -> 0.3246Hurt
Use Fourier scale-equivariant layers instead of Fourier CNNOurs · MNIST-scale idealscale-consistency: 0.2421 -> 0.9716Improved
Reduce training samples from 5000 to 1000 (ablation)Per Res. CNN · MNIST-scaleaccuracy: 0.9118 -> 0.5815Hurt
Reduce training samples from 5000 to 1000Ours · MNIST-scaleaccuracy: 0.9835 -> 0.9606Hurt
Train only on full-resolution MNIST (unseen scales)Ours · MNIST-scale missing scalesaccuracy: 0.9888 vs 0.9914No clear effect

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