Neural Canonical Transformation with Symplectic Flows

Abstract

Canonical transformation plays a fundamental role in simplifying and solving classical Hamiltonian systems. We construct flexible and powerful canonical transformations as generative models using symplectic neural networks. The model transforms physical variables towards a latent representation with an independent harmonic oscillator Hamiltonian. Correspondingly, the phase space density of the physical system flows towards a factorized Gaussian distribution in the latent space. Since the canonical transformation preserves the Hamiltonian evolution, the model captures nonlinear collective modes in the learned latent representation. We present an efficient implementation of symplectic neural coordinate transformations and two ways to train the model. The variational free energy calculation is based on the analytical form of physical Hamiltonian. While the phase space density estimation only requires samples in the coordinate space for separable Hamiltonians. We demonstrate appealing features of neural canonical transformation using toy problems including two-dimensional ring potential and harmonic chain. Finally, we apply the approach to real-world problems such as identifying slow collective modes in alanine dipeptide and conceptual compression of the MNIST dataset.

arXiv commentMain text: 9 pages, 8 figures. Supplement: 2 page, 1 figure. GitHub link: https://github.com/li012589/neuralCT

References in corpus (43)

• Adam: A Method for Stochastic Optimization
• Auto-Encoding Variational Bayes
• Escaping free-energy minima
• Generative Adversarial Networks
• Estimating Mutual Information
• Deep Potential Molecular Dynamics: a scalable model with the accuracy of quantum mechanics
• Quantum-Chemical Insights from Deep Tensor Neural Networks
• Deep learning for universal linear embeddings of nonlinear dynamics
• NICE: Non-linear Independent Components Estimation
• Variational Inference with Normalizing Flows
• NIPS 2016 Tutorial: Generative Adversarial Networks
• Machine Learning of Accurate Energy-Conserving Molecular Force Fields
• Identification of slow molecular order parameters for Markov model construction
• Density estimation using Real NVP
• VAMPnets: Deep learning of molecular kinetics
• Hamiltonian Neural Networks
• Time-lagged autoencoders: Deep learning of slow collective variables for molecular kinetics
• Markov Chain Monte Carlo and Variational Inference: Bridging the Gap
• A Variational Approach to Enhanced Sampling and Free Energy Calculations
• Normalizing Flows for Probabilistic Modeling and Inference
• Parallel WaveNet: Fast High-Fidelity Speech Synthesis
• Data-driven model reduction and transfer operator approximation
• Deep Potential: a general representation of a many-body potential energy surface
• Self-Learning Monte Carlo Method
• Accelerate Monte Carlo Simulations with Restricted Boltzmann Machines
• Solving Statistical Mechanics Using Variational Autoregressive Networks
• Flow-based generative models for Markov chain Monte Carlo in lattice field theory
• Variational Encoding of Complex Dynamics
• Neural Network Renormalization Group
• Symplectic ODE-Net: Learning Hamiltonian Dynamics with Control
• Reinforced dynamics for enhanced sampling in large atomic and molecular systems
• Towards Conceptual Compression
• Hamiltonian Graph Networks with ODE Integrators
• A-NICE-MC: Adversarial Training for MCMC
• Markov State Models from short non-Equilibrium Simulations - Analysis and Correction of Estimation Bias
• Symplectic Recurrent Neural Networks
• Physical Symmetries Embedded in Neural Networks
• Equivariant Flows: sampling configurations for multi-body systems with symmetric energies
• Equivariant Hamiltonian Flows
• Monge-Ampère Flow for Generative Modeling
• Trivializations for Gradient-Based Optimization on Manifolds
• Learning Symmetries of Classical Integrable Systems
• Self-Supervised Learning of Generative Spin-Glasses with Normalizing Flows

Cited by in corpus (21)

• Machine Learning in Nuclear Physics
• Equivariant flow-based sampling for lattice gauge theory
• Sampling using $SU(N)$ gauge equivariant flows
• Flow-matching -- efficient coarse-graining of molecular dynamics without forces
• Multiscale reweighted stochastic embedding (MRSE): Deep learning of collective variables for enhanced sampling
• Differentiable Molecular Simulations for Control and Learning
• Ab-initio study of interacting fermions at finite temperature with neural canonical transformation
• Machine-Learning Non-Conservative Dynamics for New-Physics Detection
• Symplectic Gaussian Process Regression of Hamiltonian Flow Maps
• Deep Hamiltonian networks based on symplectic integrators
• Neural canonical transformations for vibrational spectra of molecules
• Bounding free energy difference with flow matching
• Solving Quantum Statistical Mechanics with Variational Autoregressive Networks and Quantum Circuits
• E(n) Equivariant Normalizing Flows
• On Computational Poisson Geometry II: Numerical Methods
• SympNets: Intrinsic structure-preserving symplectic networks for identifying Hamiltonian systems
• Neural Network Approach to Construction of Classical Integrable Systems
• Time-Reversal Symmetric ODE Network
• Learning Non-linear Wavelet Transformation via Normalizing Flow
• Embedded-model flows: Combining the inductive biases of model-free deep learning and explicit probabilistic modeling
• Neural Canonical Transformation for the Spectra of Fluxional Molecule CH5+