Hamiltonian Dynamics with Non-Newtonian Momentum for Rapid Sampling

TL;DR

This paper introduces a novel deterministic Hamiltonian dynamics method with non-Newtonian momentum that efficiently samples from complex distributions without stochastic steps, improving convergence speed over traditional MCMC methods.

Contribution

It proposes the Energy Sampling Hamiltonian (ESH) dynamics, a new deterministic approach that enables faster, stable sampling and training of neural energy models without the need for auxiliary neural networks.

Findings

ESH dynamics converge faster than MCMC methods

The method can be interpreted as a normalizing flow without training

Specialized ODE solver improves performance significantly

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.