Parallel-tempered Stochastic Gradient Hamiltonian Monte Carlo for Approximate Multimodal Posterior Sampling
Machine Learning
2018-12-10 v2 Machine Learning
Abstract
We propose a new sampler that integrates the protocol of parallel tempering with the Nos\'e-Hoover (NH) dynamics. The proposed method can efficiently draw representative samples from complex posterior distributions with multiple isolated modes in the presence of noise arising from stochastic gradient. It potentially facilitates deep Bayesian learning on large datasets where complex multimodal posteriors and mini-batch gradient are encountered.
Cite
@article{arxiv.1812.01181,
title = {Parallel-tempered Stochastic Gradient Hamiltonian Monte Carlo for Approximate Multimodal Posterior Sampling},
author = {Rui Luo and Qiang Zhang and Yuanyuan Liu},
journal= {arXiv preprint arXiv:1812.01181},
year = {2018}
}