On Learning Hamiltonian Systems from Data

Abstract

Concise, accurate descriptions of physical systems through their conserved quantities abound in the natural sciences. In data science, however, current research often focuses on regression problems, without routinely incorporating additional assumptions about the system that generated the data. Here, we propose to explore a particular type of underlying structure in the data: Hamiltonian systems, where an "energy" is conserved. Given a collection of observations of such a Hamiltonian system over time, we extract phase space coordinates and a Hamiltonian function of them that acts as the generator of the system dynamics. The approach employs an autoencoder neural network component to estimate the transformation from observations to the phase space of a Hamiltonian system. An additional neural network component is used to approximate the Hamiltonian function on this constructed space, and the two components are trained jointly. As an alternative approach, we also demonstrate the use of Gaussian processes for the estimation of such a Hamiltonian. After two illustrative examples, we extract an underlying phase space as well as the generating Hamiltonian from a collection of movies of a pendulum. The approach is fully data-driven, and does not assume a particular form of the Hamiltonian function.

References in corpus (5)

• Deep learning for universal linear embeddings of nonlinear dynamics
• Hidden Physics Models: Machine Learning of Nonlinear Partial Differential Equations
• Invariant Variation Problems
• Neural Ordinary Differential Equations
• Inferring solutions of differential equations using noisy multi-fidelity data

Cited by in corpus (34)

• Hamiltonian neural networks for solving equations of motion
• Deep learning of thermodynamics-aware reduced-order models from data
• Structure-preserving neural networks
• Physics enhanced neural networks predict order and chaos
• Adaptable Hamiltonian neural networks
• Learning Hamiltonian dynamics by reservoir computer
• Parsimonious Physics-Informed Random Projection Neural Networks for Initial-Value Problems of ODEs and index-1 DAEs
• Fast neural Poincaré maps for toroidal magnetic fields
• Symplectic integration of learned Hamiltonian systems
• Lagrangian operator inference enhanced with structure-preserving machine learning for nonintrusive model reduction of mechanical systems
• Symplectic Gaussian Process Regression of Hamiltonian Flow Maps
• Port-metriplectic neural networks: thermodynamics-informed machine learning of complex physical systems
• Gaussian Process Port-Hamiltonian Systems: Bayesian Learning with Physics Prior
• Constrained Hamiltonian Systems and Physics-Informed Neural Networks: Hamilton-Dirac Neural Networks
• Deep Learning of Conjugate Mappings
• Variational Learning of Euler-Lagrange Dynamics from Data
• Tomography of time-dependent quantum spin networks with machine learning
• Locally-symplectic neural networks for learning volume-preserving dynamics
• Optimal criteria and their asymptotic form for data selection in data-driven reduced-order modeling with Gaussian process regression
• SafEDMD: A Koopman-based data-driven controller design framework for nonlinear dynamical systems
• Hamiltonian Neural Networks with Automatic Symmetry Detection
• Extracting Dynamical Frequencies from Invariants of Motion in Finite-Dimensional Nonlinear Integrable Systems
• Machine-learning Kohn-Sham potential from dynamics in time-dependent Kohn-Sham systems
• Generating extreme quantum scattering in graphene with machine learning
• Optimal parent Hamiltonians for time-dependent states
• Learning the Latent dynamics of Fluid flows from High-Fidelity Numerical Simulations using Parsimonious Diffusion Maps
• Quantum Process Tomography of Unitary Maps from Time-Delayed Measurements
• Learning of discrete models of variational PDEs from data
• Learning discrete Lagrangians for variational PDEs from data and detection of travelling waves
• Kernel-based parameter estimation of dynamical systems with unknown observation functions
• Machine learning of continuous and discrete variational ODEs with convergence guarantee and uncertainty quantification
• What's Next? Predicting Hamiltonian Dynamics from Discrete Observations of a Vector Field
• Machine learning and serving of discrete field theories -- when artificial intelligence meets the discrete universe
• NySALT: Nyström-type inference-based schemes adaptive to large time-stepping