On the Universality of Rotation Equivariant Point Cloud Networks
arXiv:2010.02449
Abstract
Learning functions on point clouds has applications in many fields, including computer vision, computer graphics, physics, and chemistry. Recently, there has been a growing interest in neural architectures that are invariant or equivariant to all three shape-preserving transformations of point clouds: translation, rotation, and permutation. In this paper, we present a first study of the approximation power of these architectures. We first derive two sufficient conditions for an equivariant architecture to have the universal approximation property, based on a novel characterization of the space of equivariant polynomials. We then use these conditions to show that two recently suggested models are universal, and for devising two other novel universal architectures.
References in corpus (12)
- Tensor field networks: Rotation- and translation-equivariant neural networks for 3D point clouds
- SE(3)-Transformers: 3D Roto-Translation Equivariant Attention Networks
- Invariant and Equivariant Graph Networks
- Covariant Compositional Networks For Learning Graphs
- Lorentz Group Equivariant Neural Network for Particle Physics
- N-body Networks: a Covariant Hierarchical Neural Network Architecture for Learning Atomic Potentials
- On Learning Sets of Symmetric Elements
- Set2Graph: Learning Graphs From Sets
- A Simple Proof of the Universality of Invariant/Equivariant Graph Neural Networks
- Universal Equivariant Multilayer Perceptrons
- Quaternion Equivariant Capsule Networks for 3D Point Clouds
- On Universal Equivariant Set Networks
Cited by in corpus (6)
- A Practical Method for Constructing Equivariant Multilayer Perceptrons for Arbitrary Matrix Groups
- Machine learning and invariant theory
- Coordinate Independent Convolutional Networks -- Isometry and Gauge Equivariant Convolutions on Riemannian Manifolds
- Frame Averaging for Invariant and Equivariant Network Design
- Complete Neural Networks for Complete Euclidean Graphs
- Equivariant Networks for Pixelized Spheres