Generalizing Convolutional Neural Networks for Equivariance to Lie Groups on Arbitrary Continuous Data

Abstract

The translation equivariance of convolutional layers enables convolutional neural networks to generalize well on image problems. While translation equivariance provides a powerful inductive bias for images, we often additionally desire equivariance to other transformations, such as rotations, especially for non-image data. We propose a general method to construct a convolutional layer that is equivariant to transformations from any specified Lie group with a surjective exponential map. Incorporating equivariance to a new group requires implementing only the group exponential and logarithm maps, enabling rapid prototyping. Showcasing the simplicity and generality of our method, we apply the same model architecture to images, ball-and-stick molecular data, and Hamiltonian dynamical systems. For Hamiltonian systems, the equivariance of our models is especially impactful, leading to exact conservation of linear and angular momentum.

arXiv commentICML 2020. Code available at https://github.com/mfinzi/LieConv

References in corpus (19)

• Adam: A Method for Stochastic Optimization
• Neural Message Passing for Quantum Chemistry
• Fast and Accurate Modeling of Molecular Atomization Energies with Machine Learning
• SchNet - a deep learning architecture for molecules and materials
• Wide Residual Networks
• Searching for Activation Functions
• Interaction Networks for Learning about Objects, Relations and Physics
• Group Equivariant Convolutional Networks
• Tensor field networks: Rotation- and translation-equivariant neural networks for 3D point clouds
• Hamiltonian Neural Networks
• On the Generalization of Equivariance and Convolution in Neural Networks to the Action of Compact Groups
• 3D Steerable CNNs: Learning Rotationally Equivariant Features in Volumetric Data
• Rotation equivariant vector field networks
• Data-Efficient Reinforcement Learning with Probabilistic Model Predictive Control
• Symplectic ODE-Net: Learning Hamiltonian Dynamics with Control
• General $E(2)$-Equivariant Steerable CNNs
• Hamiltonian Graph Networks with ODE Integrators
• Polar Transformer Networks
• B-Spline CNNs on Lie Groups

Cited by in corpus (28)

• A Comprehensive Survey on Deep Graph Representation Learning
• Discovering Symbolic Models from Deep Learning with Inductive Biases
• Evaluation of the MACE Force Field Architecture: from Medicinal Chemistry to Materials Science
• E(n) Equivariant Graph Neural Networks
• Incorporating Symmetry into Deep Dynamics Models for Improved Generalization
• A Survey of Deep Learning for Scientific Discovery
• Port-Hamiltonian Neural Networks for Learning Explicit Time-Dependent Dynamical Systems
• Toward Decoding the Relationship between Domain Structure and Functionality in Ferroelectrics via Hidden Latent Variables
• FlexConv: Continuous Kernel Convolutions with Differentiable Kernel Sizes
• Learning Invariances in Neural Networks
• Learning Neural Event Functions for Ordinary Differential Equations
• Automatic Symmetry Discovery with Lie Algebra Convolutional Network
• Simplifying Hamiltonian and Lagrangian Neural Networks via Explicit Constraints
• Introduction to Normalizing Flows for Lattice Field Theory
• CKConv: Continuous Kernel Convolution For Sequential Data
• Generative Models as Distributions of Functions
• Coordinate Independent Convolutional Networks -- Isometry and Gauge Equivariant Convolutions on Riemannian Manifolds
• Symmetry-Aware Actor-Critic for 3D Molecular Design
• Universal Approximation Theorem for Equivariant Maps by Group CNNs
• Commutative Lie Group VAE for Disentanglement Learning
• Probabilistic Numeric Convolutional Neural Networks
• Disentangling by Subspace Diffusion
• Trajectory Prediction using Equivariant Continuous Convolution
• Group Equivariant Stand-Alone Self-Attention For Vision
• Group Equivariant Conditional Neural Processes
• Equivariant neural networks for inverse problems
• Group Equivariant Subsampling
• Exploiting Redundancy: Separable Group Convolutional Networks on Lie Groups