Hamiltonian Dynamics with Non-Newtonian Momentum for Rapid Sampling

Abstract

Sampling from an unnormalized probability distribution is a fundamental problem in machine learning with applications including Bayesian modeling, latent factor inference, and energy-based model training. After decades of research, variations of MCMC remain the default approach to sampling despite slow convergence. Auxiliary neural models can learn to speed up MCMC, but the overhead for training the extra model can be prohibitive. We propose a fundamentally different approach to this problem via a new Hamiltonian dynamics with a non-Newtonian momentum. In contrast to MCMC approaches like Hamiltonian Monte Carlo, no stochastic step is required. Instead, the proposed deterministic dynamics in an extended state space exactly sample the target distribution, specified by an energy function, under an assumption of ergodicity. Alternatively, the dynamics can be interpreted as a normalizing flow that samples a specified energy model without training. The proposed Energy Sampling Hamiltonian (ESH) dynamics have a simple form that can be solved with existing ODE solvers, but we derive a specialized solver that exhibits much better performance. ESH dynamics converge faster than their MCMC competitors enabling faster, more stable training of neural network energy models.

arXiv comment31 pages, 19 figures. Advances in Neural Information Processing Systems (NeurIPS), 2021. Animations at https://sites.google.com/view/esh-dynamics/home, code at https://github.com/gregversteeg/esh_dynamics

References in corpus (15)

• A nonequilibrium equality for free energy differences
• Deep Unsupervised Learning using Nonequilibrium Thermodynamics
• Variational Inference with Normalizing Flows
• Markov Chain Monte Carlo and Variational Inference: Bridging the Gap
• Implicit Generation and Generalization in Energy-Based Models
• Learning Non-Convergent Non-Persistent Short-Run MCMC Toward Energy-Based Model
• Your Classifier is Secretly an Energy Based Model and You Should Treat it Like One
• How to Train Your Energy-Based Models
• NeuTra-lizing Bad Geometry in Hamiltonian Monte Carlo Using Neural Transport
• Maximum Entropy Generators for Energy-Based Models
• Hamiltonian Monte Carlo Without Detailed Balance
• Logarithmic oscillators: ideal Hamiltonian thermostats
• Improved Contrastive Divergence Training of Energy Based Models
• Dynamical computation of the density of states and Bayes factors using nonequilibrium importance sampling
• No MCMC for me: Amortized sampling for fast and stable training of energy-based models