Gradient-free Hamiltonian Monte Carlo with Efficient Kernel Exponential Families

Abstract

We propose Kernel Hamiltonian Monte Carlo (KMC), a gradient-free adaptive MCMC algorithm based on Hamiltonian Monte Carlo (HMC). On target densities where classical HMC is not an option due to intractable gradients, KMC adaptively learns the target's gradient structure by fitting an exponential family model in a Reproducing Kernel Hilbert Space. Computational costs are reduced by two novel efficient approximations to this gradient. While being asymptotically exact, KMC mimics HMC in terms of sampling efficiency, and offers substantial mixing improvements over state-of-the-art gradient free samplers. We support our claims with experimental studies on both toy and real-world applications, including Approximate Bayesian Computation and exact-approximate MCMC.

20 pages, 7 figures

References in corpus (4)

• The pseudo-marginal approach for efficient Monte Carlo computations
• Optimizing The Integrator Step Size for Hamiltonian Monte Carlo
• Optimal Rates for Random Fourier Features
• The Fundamental Incompatibility of Hamiltonian Monte Carlo and Data Subsampling

Cited by in corpus (18)

• Kernel Mean Embedding of Distributions: A Review and Beyond
• Learning Decentralized Controllers for Robot Swarms with Graph Neural Networks
• A Kernel Test of Goodness of Fit
• Sliced Score Matching: A Scalable Approach to Density and Score Estimation
• Gradient Estimators for Implicit Models
• Optimal Rates for Random Fourier Features
• Modified Hamiltonian Monte Carlo for Bayesian inference
• Learning deep kernels for exponential family densities
• Geometric Methods for Sampling, Optimisation, Inference and Adaptive Agents
• Deep Markov Chain Monte Carlo
• Walsh-Hadamard Variational Inference for Bayesian Deep Learning
• Efficient Learning of Generative Models via Finite-Difference Score Matching
• Scalable Approximate Inference and Some Applications
• Nonparametric Score Estimators
• Maximum conditional entropy Hamiltonian Monte Carlo sampler
• A data-driven and model-based accelerated Hamiltonian Monte Carlo method for Bayesian elliptic inverse problems
• Variance reduction for Random Coordinate Descent-Langevin Monte Carlo
• Generative Particle Variational Inference via Estimation of Functional Gradients