Adaptable Hamiltonian neural networks

Abstract

The rapid growth of research in exploiting machine learning to predict chaotic systems has revived a recent interest in Hamiltonian Neural Networks (HNNs) with physical constraints defined by the Hamilton's equations of motion, which represent a major class of physics-enhanced neural networks. We introduce a class of HNNs capable of adaptable prediction of nonlinear physical systems: by training the neural network based on time series from a small number of bifurcation-parameter values of the target Hamiltonian system, the HNN can predict the dynamical states at other parameter values, where the network has not been exposed to any information about the system at these parameter values. The architecture of the HNN differs from the previous ones in that we incorporate an input parameter channel, rendering the HNN parameter--cognizant. We demonstrate, using paradigmatic Hamiltonian systems, that training the HNN using time series from as few as four parameter values bestows the neural machine with the ability to predict the state of the target system in an entire parameter interval. Utilizing the ensemble maximum Lyapunov exponent and the alignment index as indicators, we show that our parameter-cognizant HNN can successfully predict the route of transition to chaos. Physics-enhanced machine learning is a forefront area of research, and our adaptable HNNs provide an approach to understanding machine learning with broad applications.

arXiv comment12 pages, 9 figures

References in corpus (22)

• Machine learning and the physical sciences
• Solving the Quantum Many-Body Problem with Artificial Neural Networks
• Using Machine Learning to Replicate Chaotic Attractors and Calculate Lyapunov Exponents from Data
• Discovering physical concepts with neural networks
• Deep Learning and its Application to LHC Physics
• Attractor Reconstruction by Machine Learning
• Multilayer multi-configuration time-dependent Hartree method: implementation and applications to a Henon-Heiles Hamiltonian and to pyrazine
• Hybrid Forecasting of Chaotic Processes: Using Machine Learning in Conjunction with a Knowledge-Based Model
• Forecasting Chaotic Systems with Very Low Connectivity Reservoir Computers
• Reservoir computing with a single time-delay autonomous Boolean node
• On Learning Hamiltonian Systems from Data
• Using reservoir computers to distinguish chaotic signals
• Machine-learning inference of fluid variables from data using reservoir computing
• Deep Lagrangian Networks: Using Physics as Model Prior for Deep Learning
• Physics enhanced neural networks predict order and chaos
• Hamiltonian Graph Networks with ODE Integrators
• Lagrangian Neural Networks
• Dynamical learning of dynamics
• Geometry of dynamics and phase transitions in classical lattice phi^4 theories
• Classifying orbits in the classical Henon-Heiles Hamiltonian system
• Inferring the dynamics of oscillatory systems using recurrent neural networks
• Teaching Recurrent Neural Networks to Modify Chaotic Memories by Example

Cited by in corpus (10)

• Hamiltonian neural networks for solving equations of motion
• Learning Hamiltonian dynamics by reservoir computer
• Machine Learning Potential of a Single Pendulum
• Tomography of time-dependent quantum spin networks with machine learning
• Learning unidirectional coupling using echo-state network
• Machine-learning Kohn-Sham potential from dynamics in time-dependent Kohn-Sham systems
• Generating extreme quantum scattering in graphene with machine learning
• Neuronal diversity can improve machine learning for physics and beyond
• Hamiltonian Neural Networks approach to fuzzball geodesics
• Deep learning of thermodynamic laws from microscopic dynamics