NAM: Normalization-based Attention Module

yichao liu, Zongru Shao, yueyang Teng, Nico Hoffmann

Abstract: Recognizing less salient features is the key for model compression. However, it has not been investigated in the revolutionary attention mechanisms. In this work, we propose a novel normalization-based attention module (NAM), which suppresses less salient weights. It applies a weight sparsity penalty to the attention modules, thus, making them more computational efficient while retaining similar performance. A comparison with three other attention mechanisms on both Resnet and Mobilenet indicates that our method results in higher accuracy. Code for this paper can be publicly accessed at \url{https://github.com/Christian-lyc/NAM}.

Submission Track: Extended abstract track, 3 pages max

Community Implementations: [ 1 code implementation](https://www.catalyzex.com/paper/arxiv:2111.12419/code)