A Novel Convolutional Neural Network Architecture with a Continuous Symmetry

Yao Liu; Hang Shao; Bing Bai

A Novel Convolutional Neural Network Architecture with a Continuous Symmetry

Computer Vision and Pattern Recognition 2024-05-21 v4 Machine Learning Neural and Evolutionary Computing

Authors: Yao Liu , Hang Shao , Bing Bai

Abstract

This paper introduces a new Convolutional Neural Network (ConvNet) architecture inspired by a class of partial differential equations (PDEs) called quasi-linear hyperbolic systems. With comparable performance on the image classification task, it allows for the modification of the weights via a continuous group of symmetry. This is a significant shift from traditional models where the architecture and weights are essentially fixed. We wish to promote the (internal) symmetry as a new desirable property for a neural network, and to draw attention to the PDE perspective in analyzing and interpreting ConvNets in the broader Deep Learning community.

Keywords

equivariant neural network convolutional neural network deep neural networks

Cite

@article{arxiv.2308.01621,
  title  = {A Novel Convolutional Neural Network Architecture with a Continuous Symmetry},
  author = {Yao Liu and Hang Shao and Bing Bai},
  journal= {arXiv preprint arXiv:2308.01621},
  year   = {2024}
}

Comments

Accepted by the 3rd CAAI International Conference on Artificial Intelligence (CICAI), 2023; with Addendum + minor edits

A Novel Convolutional Neural Network Architecture with a Continuous Symmetry

Abstract

Keywords

Cite

Comments

Related papers