reSGLD

Replica exchange stochastic gradient Langevin Dynamics

GeneralIntroduced 20003 papers
Source Paper

Description

reSGLD proposes to simulate a high-temperature particle for exploration and a low-temperature particle for exploitation and allows them to swap simultaneously. Moreover, a correction term is included to avoid biases.

Papers Using This Method

Constrained Exploration via Reflected Replica Exchange Stochastic Gradient Langevin Dynamics2024-05-13A Contour Stochastic Gradient Langevin Dynamics Algorithm for Simulations of Multi-modal Distributions2020-10-19Non-convex Learning via Replica Exchange Stochastic Gradient MCMC2020-08-12