Nonlinear Hamiltonian Monte Carlo & its Particle Approximation

2023/08/22 by Nawaf Bou‐Rabee, Bou-Rabee, Nawaf, Katharina Schuh +1 · 1 citation
Mathematics · Computer Science · #Markov Chains and Monte Carlo Methods #Mathematical Approximation and Integration #Stochastic Gradient Optimization Techniques

paper · pdf · doi:10.48550/arxiv.2308.11491

Abstract

We present a nonlinear (in the sense of McKean) generalization of Hamiltonian Monte Carlo (HMC) termed nonlinear HMC (nHMC) capable of sampling from nonlinear probability measures of mean-field type. When the underlying confinement potential is K-strongly convex and L-gradient Lipschitz, and the underlying interaction potential is gradient Lipschitz, nHMC can produce an ε-accurate approximation of a d-dimensional nonlinear probability measure in L¹-Wasserstein distance using O((L/K) log(1/ε)) steps. Owing to a uniform-in-steps propagation of chaos phenomenon, and without further regularity assumptions, unadjusted HMC with randomized time integration for the corresponding particle approximation can achieve ε-accuracy in L¹-Wasserstein distance using O( (L/K)^5/3 (d/K)^4/3 (1/ε)^8/3 log(1/ε) ) gradient evaluations. These mixing/complexity upper bounds are a specific case of more general results developed in the paper for a larger class of non-logconcave, nonlinear probability measures of mean-field type.

Cited by

• Analysis of Langevin midpoint methods using an anticipative Girsanov theorem

Related

• Rapid Mixing of Hamiltonian Monte Carlo on Strongly Log-Concave Distributions
• Randomized Hamiltonian Monte Carlo as Scaling Limit of the Bouncy Particle Sampler and Dimension-Free Convergence Rates
• Hamiltonian Monte Carlo vs. event-chain Monte Carlo: an appraisal of sampling strategies beyond the diffusive regime
• On nonlinear Markov chain Monte Carlo
• Coupling and Convergence for Hamiltonian Monte Carlo