Data Augmentation and Regularization for Learning Group Equivariance

TL;DR

This paper explores how data augmentation combined with regularization can effectively teach machine learning models to learn group equivariance, leveraging known symmetries to enhance performance.

Contribution

It extends previous work by demonstrating that training with augmented data and regularization can achieve model equivariance, providing a practical approach.

Findings

Equivariance can be learned through combined data augmentation and regularization.

Training with augmented data improves model symmetry properties.

The approach generalizes previous results on learning group equivariance.

Abstract

In many machine learning tasks, known symmetries can be used as an inductive bias to improve model performance. In this paper, we consider learning group equivariance through training with data augmentation. We summarize results from a previous paper of our own, and extend the results to show that equivariance of the trained model can be achieved through training on augmented data in tandem with regularization.