CVLGOct 4, 2021

Improving Axial-Attention Network Classification via Cross-Channel Weight Sharing

arXiv:2110.01185v2
Originality Synthesis-oriented
AI Analysis

This incremental approach may enhance classification accuracy for a broad range of networks, particularly in image processing tasks.

The study replaced layers in an Axial Attention network with representationally coherent variants to improve image classification, resulting in improved accuracy on the ImageNet300k dataset compared to baseline networks.

In recent years, hypercomplex-inspired neural networks (HCNNs) have been used to improve deep learning architectures due to their ability to enable channel-based weight sharing, treat colors as a single entity, and improve representational coherence within the layers. The work described herein studies the effect of replacing existing layers in an Axial Attention network with their representationally coherent variants to assess the effect on image classification. We experiment with the stem of the network, the bottleneck layers, and the fully connected backend, by replacing them with representationally coherent variants. These various modifications lead to novel architectures which all yield improved accuracy performance on the ImageNet300k classification dataset. Our baseline networks for comparison were the original real-valued ResNet, the original quaternion-valued ResNet, and the Axial Attention ResNet. Since improvement was observed regardless of which part of the network was modified, there is a promise that this technique may be generally useful in improving classification accuracy for a large class of networks.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes