Constrained Exploration via Reflected Replica Exchange Stochastic Gradient Langevin Dynamics

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Zheng, Haoyang, Du, Hengrong, Feng, Qi, Deng, Wei, Lin, Guang
Format: Preprint
Publié: 2024
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866909215280332800
author Zheng, Haoyang
Du, Hengrong
Feng, Qi
Deng, Wei
Lin, Guang
author_facet Zheng, Haoyang
Du, Hengrong
Feng, Qi
Deng, Wei
Lin, Guang
contents Replica exchange stochastic gradient Langevin dynamics (reSGLD) is an effective sampler for non-convex learning in large-scale datasets. However, the simulation may encounter stagnation issues when the high-temperature chain delves too deeply into the distribution tails. To tackle this issue, we propose reflected reSGLD (r2SGLD): an algorithm tailored for constrained non-convex exploration by utilizing reflection steps within a bounded domain. Theoretically, we observe that reducing the diameter of the domain enhances mixing rates, exhibiting a $\textit{quadratic}$ behavior. Empirically, we test its performance through extensive experiments, including identifying dynamical systems with physical constraints, simulations of constrained multi-modal distributions, and image classification tasks. The theoretical and empirical findings highlight the crucial role of constrained exploration in improving the simulation efficiency.
format Preprint
id arxiv_https___arxiv_org_abs_2405_07839
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Constrained Exploration via Reflected Replica Exchange Stochastic Gradient Langevin Dynamics
Zheng, Haoyang
Du, Hengrong
Feng, Qi
Deng, Wei
Lin, Guang
Machine Learning
Artificial Intelligence
Replica exchange stochastic gradient Langevin dynamics (reSGLD) is an effective sampler for non-convex learning in large-scale datasets. However, the simulation may encounter stagnation issues when the high-temperature chain delves too deeply into the distribution tails. To tackle this issue, we propose reflected reSGLD (r2SGLD): an algorithm tailored for constrained non-convex exploration by utilizing reflection steps within a bounded domain. Theoretically, we observe that reducing the diameter of the domain enhances mixing rates, exhibiting a $\textit{quadratic}$ behavior. Empirically, we test its performance through extensive experiments, including identifying dynamical systems with physical constraints, simulations of constrained multi-modal distributions, and image classification tasks. The theoretical and empirical findings highlight the crucial role of constrained exploration in improving the simulation efficiency.
title Constrained Exploration via Reflected Replica Exchange Stochastic Gradient Langevin Dynamics
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2405.07839