A Practical Method for Constructing Equivariant Multilayer Perceptrons for Arbitrary Matrix Groups

arXiv2104.09459MaRDI QIDQ6365648

Author name not available (Why is that?)

Publication date: 19 April 2021

Abstract: Symmetries and equivariance are fundamental to the generalization of neural networks on domains such as images, graphs, and point clouds. Existing work has primarily focused on a small number of groups, such as the translation, rotation, and permutation groups. In this work we provide a completely general algorithm for solving for the equivariant layers of matrix groups. In addition to recovering solutions from other works as special cases, we construct multilayer perceptrons equivariant to multiple groups that have never been tackled before, including $mathrmO(1,3)$, $mathrmO(5)$, $mathrmSp(n)$, and the Rubik's cube group. Our approach outperforms non-equivariant baselines, with applications to particle physics and dynamical systems. We release our software library to enable researchers to construct equivariant layers for arbitrary matrix groups.

Has companion code repository: https://github.com/zichunhao/lgn-autoencoder

This page was built for publication: A Practical Method for Constructing Equivariant Multilayer Perceptrons for Arbitrary Matrix Groups

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6365648)