Probing the effects of broken symmetries in machine learning

Abstract

Symmetry is one of the most central concepts in physics, and it is no surprise that it has also been widely adopted as an inductive bias for machine-learning models applied to the physical sciences. This is especially true for models targeting the properties of matter at the atomic scale. Both established and state-of-the-art approaches, with almost no exceptions, are built to be exactly equivariant to translations, permutations, and rotations of the atoms. Incorporating symmetries -- rotations in particular -- constrains the model design space and implies more complicated architectures that are often also computationally demanding. There are indications that non-symmetric models can easily learn symmetries from data, and that doing so can even be beneficial for the accuracy of the model. We put a model that obeys rotational invariance only approximately to the test, in realistic scenarios involving simulations of gas-phase, liquid, and solid water. We focus specifically on physical observables that are likely to be affected -- directly or indirectly -- by symmetry breaking, finding negligible consequences when the model is used in an interpolative, bulk, regime. Even for extrapolative gas-phase predictions, the model remains very stable, even though symmetry artifacts are noticeable. We also discuss strategies that can be used to systematically reduce the magnitude of symmetry breaking when it occurs, and assess their impact on the convergence of observables.

References in corpus (34)

• Canonical sampling through velocity-rescaling
• Gaussian Approximation Potentials: the accuracy of quantum mechanics, without the electrons
• Machine learning and the physical sciences
• Fast and Accurate Modeling of Molecular Atomization Energies with Machine Learning
• Deep Potential Molecular Dynamics: a scalable model with the accuracy of quantum mechanics
• PCT: Point cloud transformer
• E(3)-Equivariant Graph Neural Networks for Data-Efficient and Accurate Interatomic Potentials
• Physics-inspired structural representations for molecules and materials
• An Efficient and Accurate Car-Parrinello-like Approach to Born-Oppenheimer Molecular Dynamics
• i-PI 2.0: A Universal Force Engine for Advanced Molecular Simulations
• Ab initio thermodynamics of liquid and solid water
• MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fields
• The first-principles phase diagram of monolayer nanoconfined water
• Colored-noise thermostats à la carte
• On the Completeness of Atomic Structure Representations
• Forces are not Enough: Benchmark and Critical Evaluation for Machine Learning Force Fields with Molecular Simulations
• Stochastic thermostats: comparison of local and global schemes
• Representations of molecules and materials for interpolation of quantum-mechanical simulations via machine learning
• Transferable atomic multipole machine learning models for small organic molecules
• The Design Space of E(3)-Equivariant Atom-Centered Interatomic Potentials
• i-PI 3.0: a flexible and efficient framework for advanced atomistic simulations
• Atomic Permutationally Invariant Polynomials for Fitting Molecular Force Fields
• Accelerated ab-initio Molecular Dynamics: probing the weak dispersive forces in dense liquid hydrogen
• TensorNet: Cartesian Tensor Representations for Efficient Learning of Molecular Potentials
• Smooth, exact rotational symmetrization for deep learning on point clouds
• So3krates: Equivariant attention for interactions on arbitrary length-scales in molecular systems
• Black-box inhomogeneous preconditioning for self-consistent field iterations in density functional theory
• Gauss-Legendre Sampling on the Rotation Group
• FAENet: Frame Averaging Equivariant GNN for Materials Modeling
• Equivariance with Learned Canonicalization Functions
• Learning Probabilistic Symmetrization for Architecture Agnostic Equivariance
• Symmetry Breaking and Equivariant Neural Networks
• Equivariant Frames and the Impossibility of Continuous Canonicalization
• Equivariance via Minimal Frame Averaging for More Symmetries and Efficiency

Cited by in corpus (1)

• A deep learning model for chemical shieldings in molecular organic solids including anisotropy