The Selective G-Bispectrum and its Inversion: Applications to G-Invariant Networks

TL;DR

This paper introduces a computationally efficient selective G-Bispectrum for achieving group invariance in neural networks, improving accuracy and robustness while reducing complexity from quadratic to linear.

Contribution

It proposes a novel selective G-Bispectrum that reduces computational complexity and maintains desirable mathematical properties for G-invariant neural network applications.

Findings

Selective G-Bispectrum reduces computation from O(|G|^2) to O(|G|).

Integration of the selective G-Bispectrum improves neural network accuracy and robustness.

The method demonstrates significant speed-ups over traditional G-Bispectrum computations.

Abstract

An important problem in signal processing and deep learning is to achieve \textit{invariance} to nuisance factors not relevant for the task. Since many of these factors are describable as the action of a group $G$ (e.g. rotations, translations, scalings), we want methods to be $G$-invariant. The $G$-Bispectrum extracts every characteristic of a given signal up to group action: for example, the shape of an object in an image, but not its orientation. Consequently, the $G$-Bispectrum has been incorporated into deep neural network architectures as a computational primitive for $G$-invariance\textemdash akin to a pooling mechanism, but with greater selectivity and robustness. However, the computational cost of the $G$-Bispectrum ($\mathcal{O}(|G|^2)$, with $|G|$ the size of the group) has limited its widespread adoption. Here, we show that the $G$-Bispectrum computation contains redundancies that can be reduced into a \textit{selective $G$-Bispectrum} with $\mathcal{O}(|G|)$ complexity. We prove desirable mathematical properties of the selective $G$-Bispectrum and demonstrate how its integration in neural networks enhances accuracy and robustness compared to traditional approaches, while enjoying considerable speeds-up compared to the full $G$-Bispectrum.