Non-reversible parallel tempering for deep posterior approximation

The dataset used in the paper is a multi-modal distribution, and the authors propose a new method for parallel tempering with stochastic gradient Langevin dynamics.

BibTex: