Abstract:Symmetry is present throughout nature and continues to play an increasingly central role in physics and machine learning. Fundamental symmetries, such as Poincaré invariance, allow physical laws discovered in laboratories on Earth to be extrapolated to the farthest reaches of the universe. Symmetry is essential to achieving this extrapolatory power in machine learning applications. For example, translation invariance in image classification allows models with fewer parameters, such as convolutional neural networks, to be trained on smaller data sets and achieve state-of-the-art performance. In this paper, we provide a unifying theoretical and methodological framework for incorporating symmetry into machine learning models in three ways: 1. enforcing known symmetry when training a model; 2. discovering unknown symmetries of a given model or data set; and 3. promoting symmetry during training by learning a model that breaks symmetries within a user-specified group of candidates when there is sufficient evidence in the data. We show that these tasks can be cast within a common mathematical framework whose central object is the Lie derivative associated with fiber-linear Lie group actions on vector bundles. We extend and unify several existing results by showing that enforcing and discovering symmetry are linear-algebraic tasks that are dual with respect to the bilinear structure of the Lie derivative. We also propose a novel way to promote symmetry by introducing a class of convex regularization functions based on the Lie derivative and nuclear norm relaxation to penalize symmetry breaking during training of machine learning models. We explain how these ideas can be applied to a wide range of machine learning models including basis function regression, dynamical systems discovery, neural networks, and neural operators acting on fields.

Symmetry within Solutions

Symmetry within and between solutions

Symmetry of Solutions to a Class of Geometric Equations for Hypersurfaces in Rn+1

Detecting Design Intent in Approximate CAD Models Using Symmetry.

Algorithms Transcending the SAT-Symmetry Interface

Skeleton-Intrinsic Symmetrization of Shapes

Breaking Value Symmetry

Improving Discrete Model Representations via Symmetry Considerations

Symmetries in Integer Programs

A Unified Framework to Enforce, Discover, and Promote Symmetry in Machine Learning

satsuma: Structure-based Symmetry Breaking in SAT

Using Symmetries to Lift Satisfiability Checking

Symmetries in Linear and Integer Programs

An adaptive prefix-assignment technique for symmetry reduction

Static Symmetry Breaking in Many-Sorted Finite Model Finding

Symmetry properties and explicit solutions of the generalized Weierstrass system

Symmetry of Integral Equations on Bounded Domains

On the symmetry definitions for the constraint dynamical systems

Symmetry-breaking Answer Set Solving

Detecting and Handling Reflection Symmetries in Mixed-Integer (Nonlinear) Programming

Symmetries of Dependency Quantified Boolean Formulas