AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Weight Excitation: Built-in Attention Mechanisms in Convolutional Neural Networks

Niamul Quader, Md Mafijul Islam Bhuiyan, Juwei Lu, Peng Dai, Wei Li

ECCV 2020 · Architectures and efficiency › Attention and transformers · approach: Theory of self-attention dynamics · 30 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 location-based weight excitation (LWE)ResNet50 · ImageNet valtop-1 accuracy (%): 75.8 -> 77.1Improved
add magnitude-based weight excitation (MWE) (ablation)ResNet50 · ImageNet valtop-1 accuracy (%): 75.83 -> 76.52Improved
add LWE instead of MWE (ablation)ResNet50 · ImageNet valtop-1 accuracy (%): 76.52 -> 77.07Improved
combine LWE with MWE (ablation)ResNet50 · ImageNet valtop-1 accuracy (%): 77.07 -> 77.09No clear effect
add LWE on top of SE (ablation)ResNet50-SE · ImageNet valtop-1 accuracy (%): 77.1 -> 77.5Improved
add LWE on top of SE (ablation)ResNet152-SE · ImageNet valtop-1 accuracy (%): 78.7 -> 79.0Improved
add LWE to depthwise-conv networkMobileNetV2 · ImageNet valtop-1 accuracy (%): 66.2 -> 66.5Mixed
use MWE on depthwise convolutions instead of LWE (ablation)MobileNetV2 · ImageNet valtop-1 accuracy (%): 66.5 -> 67.0Improved
add LWE vs weight standardization/normalization (ablation)ResNet50 · ImageNet valtop-1 accuracy (%): 76.1 -> 77.1Improved
add spectral-norm style reparameterization (SN) (ablation)ResNet50 · ImageNet valtop-1 accuracy (%): 75.8 -> 75.7No clear effect
train LWE for 90 epochs instead of 100 (ablation)ResNet50-LWE · ImageNet valtop-1 accuracy (%): 77.1 -> 76.9No clear effect
set MWE epsilon_A to 10^-3 / 0.1 / 0.2 (ablation)ResNet50-MWE · ImageNet valtop-1 accuracy (%): 76.38 / 76.52 / 76.43No 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