Lie Neurons: Adjoint-Equivariant Neural Networks for Semisimple Lie Algebras

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Lin, Tzu-Yuan, Zhu, Minghan, Ghaffari, Maani
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910475305877504
author Lin, Tzu-Yuan
Zhu, Minghan
Ghaffari, Maani
author_facet Lin, Tzu-Yuan
Zhu, Minghan
Ghaffari, Maani
contents This paper proposes an equivariant neural network that takes data in any semi-simple Lie algebra as input. The corresponding group acts on the Lie algebra as adjoint operations, making our proposed network adjoint-equivariant. Our framework generalizes the Vector Neurons, a simple $\mathrm{SO}(3)$-equivariant network, from 3-D Euclidean space to Lie algebra spaces, building upon the invariance property of the Killing form. Furthermore, we propose novel Lie bracket layers and geometric channel mixing layers that extend the modeling capacity. Experiments are conducted for the $\mathfrak{so}(3)$, $\mathfrak{sl}(3)$, and $\mathfrak{sp}(4)$ Lie algebras on various tasks, including fitting equivariant and invariant functions, learning system dynamics, point cloud registration, and homography-based shape classification. Our proposed equivariant network shows wide applicability and competitive performance in various domains.
format Preprint
id arxiv_https___arxiv_org_abs_2310_04521
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Lie Neurons: Adjoint-Equivariant Neural Networks for Semisimple Lie Algebras
Lin, Tzu-Yuan
Zhu, Minghan
Ghaffari, Maani
Machine Learning
Artificial Intelligence
This paper proposes an equivariant neural network that takes data in any semi-simple Lie algebra as input. The corresponding group acts on the Lie algebra as adjoint operations, making our proposed network adjoint-equivariant. Our framework generalizes the Vector Neurons, a simple $\mathrm{SO}(3)$-equivariant network, from 3-D Euclidean space to Lie algebra spaces, building upon the invariance property of the Killing form. Furthermore, we propose novel Lie bracket layers and geometric channel mixing layers that extend the modeling capacity. Experiments are conducted for the $\mathfrak{so}(3)$, $\mathfrak{sl}(3)$, and $\mathfrak{sp}(4)$ Lie algebras on various tasks, including fitting equivariant and invariant functions, learning system dynamics, point cloud registration, and homography-based shape classification. Our proposed equivariant network shows wide applicability and competitive performance in various domains.
title Lie Neurons: Adjoint-Equivariant Neural Networks for Semisimple Lie Algebras
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2310.04521