AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

AttnZero: Efficient Attention Discovery for Vision Transformers

Lujun Li, Zimian Wei, Peijie Dong, Wenhan Luo, Wei Xue, Qifeng Liu, Yike Guo

ECCV 2024 · Architectures and efficiency › Attention and transformers · approach: Efficient and robust vision transformers · 14 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 softmax attention with AttnZero-discovered linear attentionDeiT-T · ImageNet-1k (224x224)Top-1 accuracy (%): 72.2 -> 74.9Improved
replace self-attention with AttnZero in larger or stronger backbonesSwin-S, Swin-B, CSwin-T · ImageNet-1k (224x224)Top-1 accuracy (%): 83.0 -> 83.2 (Swin-S); 83.5 -> 83.7 (Swin-B); 82.7 -> 82.9 (CSwin-T)Improved
use AttnZero search instead of random search in the same search space (ablation)DeiT-T · CIFAR-100Top-1 accuracy (%): 76.04 -> 77.68Improved
use multi-objective search over several ViTs instead of single-objective search (ablation)AttnZero search · ImageNet-1kTop-1 accuracy (%): +0.9%Improved
search attention on one ViT (Swin) only (ablation)AttnZero search · DeiT, PVT transferTop-1 accuracyHurt
add program checking and rejection protocols to the search (ablation)AttnZero search · search efficiencysearch speedup: more than 5x search accelerationImproved
use advanced normalization and activation operators (vs new binary operators) (ablation)search space · Attn-Bench-101 (AutoFormer, CIFAR-100)Top-1 accuracyImproved
use computation graphs i3n302 and i3n303 (ablation)search space · Attn-Bench-101 (AutoFormer, CIFAR-100)Top-1 accuracyImproved
use other linear attentions (Hydra, Efficient Attn, Linear Angular) in place of softmaxDeiT-T · ImageNet-1k (224x224)Top-1 accuracy (%): 72.2 -> 68.3 (Hydra)Hurt
replace attention with AttnZero in PVT-T for detectionPVT-T Mask R-CNN · COCO (Mask R-CNN 1x)box AP: 36.7 -> 40.0Improved
replace attention with AttnZero in PVT-T for segmentationPVT-T SemanticFPN · ADE20KmIoU: 36.57 -> 39.16Improved

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