Convergence Error Analysis of Reflected Gradient Langevin Dynamics for Globally Optimizing Non-Convex Constrained Problems

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Sato, Kanji, Takeda, Akiko, Kawai, Reiichiro, Suzuki, Taiji
Natura: Preprint
Pubblicazione: 2022
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909286325551104
author Sato, Kanji
Takeda, Akiko
Kawai, Reiichiro
Suzuki, Taiji
author_facet Sato, Kanji
Takeda, Akiko
Kawai, Reiichiro
Suzuki, Taiji
contents Gradient Langevin dynamics and a variety of its variants have attracted increasing attention owing to their convergence towards the global optimal solution, initially in the unconstrained convex framework while recently even in convex constrained non-convex problems. In the present work, we extend those frameworks to non-convex problems on a non-convex feasible region with a global optimization algorithm built upon reflected gradient Langevin dynamics and derive its convergence rates. By effectively making use of its reflection at the boundary in combination with the probabilistic representation for the Poisson equation with the Neumann boundary condition, we present promising convergence rates, particularly faster than the existing one for convex constrained non-convex problems.
format Preprint
id arxiv_https___arxiv_org_abs_2203_10215
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle Convergence Error Analysis of Reflected Gradient Langevin Dynamics for Globally Optimizing Non-Convex Constrained Problems
Sato, Kanji
Takeda, Akiko
Kawai, Reiichiro
Suzuki, Taiji
Optimization and Control
Probability
Machine Learning
Gradient Langevin dynamics and a variety of its variants have attracted increasing attention owing to their convergence towards the global optimal solution, initially in the unconstrained convex framework while recently even in convex constrained non-convex problems. In the present work, we extend those frameworks to non-convex problems on a non-convex feasible region with a global optimization algorithm built upon reflected gradient Langevin dynamics and derive its convergence rates. By effectively making use of its reflection at the boundary in combination with the probabilistic representation for the Poisson equation with the Neumann boundary condition, we present promising convergence rates, particularly faster than the existing one for convex constrained non-convex problems.
title Convergence Error Analysis of Reflected Gradient Langevin Dynamics for Globally Optimizing Non-Convex Constrained Problems
topic Optimization and Control
Probability
Machine Learning
url https://arxiv.org/abs/2203.10215