AI Paper Atlas

Research map / Architectures and efficiency / CNN design

Butterfly Transform: An Efficient FFT Based Neural Architecture Design

Keivan Alizadeh vahid, Anish Prabhu, Ali Farhadi, Mohammad Rastegari

CVPR 2020 · Architectures and efficiency › CNN design · approach: Invertible and wide-network analysis · 42 citations (OpenAlex)

Check 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 pointwise convolutions with BFTShuffleNetV2 · ImageNet validationTop-1 accuracy: 50.86 -> 55.26Improved
Replace pointwise convolutions with BFTMobileNetV1 · ImageNet validationTop-1 accuracy: 41.50 -> 46.58Improved
Replace pointwise convolutions with BFTMobileNetV3 · ImageNet validationTop-1 accuracy: 49.8 -> 55.21Improved
Use BFT channel fusion instead of standard pointwise convolutionsShuffleNetV2 at low FLOPs · ImageNet validationTop-1 accuracy: 55.26 vs 50.2; 55.26 vs 49.3Improved
Use BFT channel fusion instead of alternative structured matricesMobileNetV1 · ImageNet validationTop-1 accuracy: 54.3 vs 49.65Improved
Add a first-to-last residual connection across Butterfly Layers (ablation)MobileNetV1+BFT · CIFAR-10Accuracy: 79.2 -> 81.75Improved
Use first-to-last rather than every-other-layer residual connections (ablation)MobileNetV1+BFT · CIFAR-10Accuracy: 81.12 -> 81.75Improved
Add nonlinear activations inside Butterfly Layers (ablation)BFT · ImageNet classificationAccuracyHurt
Add batch normalization inside each Butterfly Layer (ablation)BFT · ImageNet classificationFLOPsHurt
Apply weight decay to BFT weights (ablation)BFT · ImageNet classificationAccuracyHurt
Use butterfly base k=4 rather than k=2, k=8, or k=n (ablation)MobileNetV1+BFT · ImageNet, approximately 40M FLOPsAccuracyImproved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

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