Constructing 3D Rotational Invariance and Equivariance with Symmetric Tensor Networks

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Meng, Wang, Chao, Zhang, Hao, Dong, Shaojun, He, Lixin
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915768088657920
author Zhang, Meng
Wang, Chao
Zhang, Hao
Dong, Shaojun
He, Lixin
author_facet Zhang, Meng
Wang, Chao
Zhang, Hao
Dong, Shaojun
He, Lixin
contents Symmetry-aware architectures are central to geometric deep learning. We present a systematic approach for constructing continuous rotationally invariant and equivariant functions using symmetric tensor networks. The proposed framework supports inputs and outputs given as a tuple of Cartesian tensors of different rank as well as spherical tensors of different type. We introduce tensor network generators for invariant maps and obtain equivariant maps via differentiation. Specifically, we derive general continuous equivariant maps from vector inputs to Cartesian or spherical tensor output. Finally, we clarify how common equivariant primitives in geometric graph neural networks arise within our construction.
format Preprint
id arxiv_https___arxiv_org_abs_2508_12596
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Constructing 3D Rotational Invariance and Equivariance with Symmetric Tensor Networks
Zhang, Meng
Wang, Chao
Zhang, Hao
Dong, Shaojun
He, Lixin
Machine Learning
Symmetry-aware architectures are central to geometric deep learning. We present a systematic approach for constructing continuous rotationally invariant and equivariant functions using symmetric tensor networks. The proposed framework supports inputs and outputs given as a tuple of Cartesian tensors of different rank as well as spherical tensors of different type. We introduce tensor network generators for invariant maps and obtain equivariant maps via differentiation. Specifically, we derive general continuous equivariant maps from vector inputs to Cartesian or spherical tensor output. Finally, we clarify how common equivariant primitives in geometric graph neural networks arise within our construction.
title Constructing 3D Rotational Invariance and Equivariance with Symmetric Tensor Networks
topic Machine Learning
url https://arxiv.org/abs/2508.12596