Exact conservation laws for neural network integrators of dynamical systems

DOI10.1016/J.JCP.2023.112234MaRDI QIDQ6105087zbMATH OpenOpenAlexFDO

Authors Eike H. Müller

Publication date 16 June 2023

Published in Journal of Computational Physics (Search for Journal in Brave)

Copyright license Creative Commons Attribution 4.0 International

Full work available at URL https://arxiv.org/abs/2209.11661

zbMATH Keywords

dynamical systemLagrangian mechanicsmachine learningNoether's theorem

Mathematics Subject Classification ID

General theory for ordinary differential equations (34Axx) Dynamical aspects of finite-dimensional Hamiltonian and Lagrangian systems (37Jxx) Numerical methods for ordinary differential equations (65Lxx)

Abstract: The solution of time dependent differential equations with neural networks has attracted a lot of attention recently. The central idea is to learn the laws that govern the evolution of the solution from data, which might be polluted with random noise. However, in contrast to other machine learning applications, usually a lot is known about the system at hand. For example, for many dynamical systems physical quantities such as energy or (angular) momentum are exactly conserved. Hence, the neural network has to learn these conservation laws from data and they will only be satisfied approximately due to finite training time and random noise. In this paper we present an alternative approach which uses Noether's Theorem to inherently incorporate conservation laws into the architecture of the neural network. We demonstrate that this leads to better predictions for three model systems: the motion of a non-relativistic particle in a three-dimensional Newtonian gravitational potential, the motion of a massive relativistic particle in the Schwarzschild metric and a system of two interacting particles in four dimensions.

Recommendations

• Neural network representation of time integrators
• Structure-preserving neural networks
• Locally-symplectic neural networks for learning volume-preserving dynamics
• Deep Hamiltonian neural networks based on symplectic integrators

Cites work

• Group theory. Birdtracks, Lie's, and exceptional groups
• scientific article; zbMATH DE number 2003146 (Why is no real title available?)
• Machine learning strategies for systems with invariance properties
• Solving Ordinary Differential Equations I
• Structure-preserving neural networks

Cited in

(6)

• Structure-preserving neural networks
• Machine learning post-Minkowskian integrals
• A hybrid approach for solving the gravitational N-body problem with artificial neural networks
• Deep learning of first-order nonlinear hyperbolic conservation law solvers
• ECLEIRS: exact conservation law embedded identification of reduced states for parameterized nonlinear conservation laws from sparse and noisy data
• Learning efficient and provably convergent symmetric splitting methods

This page was built for publication: Exact conservation laws for neural network integrators of dynamical systems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6105087)