Sampling by Divergence Minimization
arXiv:2105.00520 · doi:10.1080/03610918.2023.2199352
Abstract
We introduce a Markov Chain Monte Carlo (MCMC) method that is designed to sample from target distributions with irregular geometry using an adaptive scheme. In cases where targets exhibit non-Gaussian behaviour, we propose that adaption should be regional rather than global. Our algorithm minimizes the information projection component of the Kullback-Leibler (KL) divergence between the proposal and target distributions to encourage proposals that are distributed similarly to the regional geometry of the target. Unlike traditional adaptive MCMC, this procedure rapidly adapts to the geometry of the target's current position as it explores the surrounding space without the need for many preexisting samples. The divergence minimization algorithms are tested on target distributions with irregularly shaped modes and we provide results demonstrating the effectiveness of our methods.
arXiv comment33 pages, 12 figures
References in corpus (8)
- MCMC using Hamiltonian dynamics
- Improved Adaptive Rejection Metropolis Sampling Algorithms
- A Review of Multiple Try MCMC algorithms for Signal Processing
- Weak convergence of Metropolis algorithms for non-i.i.d. target distributions
- A generalization of the Kullback-Leibler divergence and its properties
- Gradient-based Adaptive Markov Chain Monte Carlo
- Stability of adversarial Markov chains, with an application to adaptive MCMC algorithms
- How well does your sampler really work?