Abstract:Many important physical systems can be described as the evolution of a Hamiltonian system, which has the important property of being conservative, that is, energy is conserved throughout the evolution. Physics Informed Neural Networks and in particular Hamiltonian Neural Networks have emerged as a mechanism to incorporate structural inductive bias into the NN model. By ensuring physical invariances are conserved, the models exhibit significantly better sample complexity and out-of-distribution accuracy than standard NNs. Learning the Hamiltonian as a function of its canonical variables, typically position and velocity, from sample observations of the system thus becomes a critical task in system identification and long-term prediction of system behavior. However, to truly preserve the long-run physical conservation properties of Hamiltonian systems, one must use symplectic integrators for a forward pass of the system's simulation. While symplectic schemes have been used in the literature, they are thus far limited to situations when they reduce to explicit algorithms, which include the case of separable Hamiltonians or augmented non-separable Hamiltonians. We extend it to generalized non-separable Hamiltonians, and noting the self-adjoint property of symplectic integrators, we bypass computationally intensive backpropagation through an ODE solver. We show that the method is robust to noise and provides a good approximation of the system Hamiltonian when the state variables are sampled from a noisy observation. In the numerical results, we show the performance of the method concerning Hamiltonian reconstruction and conservation, indicating its particular advantage for non-separable systems.

Symplecticity of coupled Hamiltonian systems

EXACT STATIONARY SOLUTIONS OF STOCHASTICALLY AND HARMONICALLY EXCITED AND DISSIPATED INTEGRABLE HAMILTONIAN SYSTEMS

Reliability of quasi integrable generalized Hamiltonian systems

Equivalent non-linear system method for stochastically excited and dissipated partially integrable Hamiltonian systems

Non-unique Hamiltonians for Discrete Symplectic Dynamics

Infinitely Many Homoclinic Orbits for Superlinear Hamiltonian Systems

Finite Dimensional Hamiltonian System Related to a Lax Pair with Symplectic and Cyclic Symmetries

On the integrability of hybrid Hamiltonian systems

Quasi-Hamiltonian Structure Associated with an Integrable Coupling System

Discrete-time port-Hamiltonian systems: A definition based on symplectic integration

Hamiltonian thermodynamics on symplectic manifolds

Generalizing Hamiltonian Mechanics with Closed Differential Forms

Duality for systems of conservation laws

Hamiltonian dynamics on the symplectic extended phase space for autonomous and non-autonomous systems

Structure-preserving quantum algorithms for linear and nonlinear Hamiltonian systems

Multidimensional Nonhomogeneous Quasi-Linear Systems and Their Hamiltonian Structure

Symplectic structure of post-Newtonian Hamiltonian for spinning compact binaries

Learning Generalized Hamiltonians using fully Symplectic Mappings

First Steps Towards a Symplectic Dynamics

Infinite-dimensional port-Hamiltonian systems with a stationary interface

Global Applicability of the Symplectic Integrator Method of Hamiltonian Systems