English

A PAC-Bayesian Generalization Bound for Equivariant Networks

Machine Learning 2022-10-25 v1 Machine Learning

Abstract

Equivariant networks capture the inductive bias about the symmetry of the learning task by building those symmetries into the model. In this paper, we study how equivariance relates to generalization error utilizing PAC Bayesian analysis for equivariant networks, where the transformation laws of feature spaces are determined by group representations. By using perturbation analysis of equivariant networks in Fourier domain for each layer, we derive norm-based PAC-Bayesian generalization bounds. The bound characterizes the impact of group size, and multiplicity and degree of irreducible representations on the generalization error and thereby provide a guideline for selecting them. In general, the bound indicates that using larger group size in the model improves the generalization error substantiated by extensive numerical experiments.

Keywords

Cite

@article{arxiv.2210.13150,
  title  = {A PAC-Bayesian Generalization Bound for Equivariant Networks},
  author = {Arash Behboodi and Gabriele Cesa and Taco Cohen},
  journal= {arXiv preprint arXiv:2210.13150},
  year   = {2022}
}

Comments

41 pages, 15 figures, accepted at NeurIPS 2022

R2 v1 2026-06-28T04:20:54.094Z