Enregistré dans:
Détails bibliographiques
Auteurs principaux: Kim, Jaeyeon, Park, Chanwoo, Ozdaglar, Asuman, Diakonikolas, Jelena, Ryu, Ernest K.
Format: Preprint
Publié: 2023
Sujets:
Accès en ligne:https://arxiv.org/abs/2311.17296
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866916246452174848
author Kim, Jaeyeon
Park, Chanwoo
Ozdaglar, Asuman
Diakonikolas, Jelena
Ryu, Ernest K.
author_facet Kim, Jaeyeon
Park, Chanwoo
Ozdaglar, Asuman
Diakonikolas, Jelena
Ryu, Ernest K.
contents While first-order optimization methods are usually designed to efficiently reduce the function value $f(x)$, there has been recent interest in methods efficiently reducing the magnitude of $\nabla f(x)$, and the findings show that the two types of methods exhibit a certain symmetry. In this work, we present mirror duality, a one-to-one correspondence between mirror-descent-type methods reducing function value and reducing gradient magnitude. Using mirror duality, we obtain the dual accelerated mirror descent (dual-AMD) method that efficiently reduces $ψ^*(\nabla f(x))$, where $ψ$ is a distance-generating function and $ψ^*$ quantifies the magnitude of $\nabla f(x)$. We then apply dual-AMD to efficiently reduce $\|\nabla f(\cdot) \|_q$ for $q\in [2,\infty)$ and to efficiently compute $\varepsilon$-approximate solutions of the optimal transport problem.
format Preprint
id arxiv_https___arxiv_org_abs_2311_17296
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Mirror Duality in Convex Optimization
Kim, Jaeyeon
Park, Chanwoo
Ozdaglar, Asuman
Diakonikolas, Jelena
Ryu, Ernest K.
Optimization and Control
While first-order optimization methods are usually designed to efficiently reduce the function value $f(x)$, there has been recent interest in methods efficiently reducing the magnitude of $\nabla f(x)$, and the findings show that the two types of methods exhibit a certain symmetry. In this work, we present mirror duality, a one-to-one correspondence between mirror-descent-type methods reducing function value and reducing gradient magnitude. Using mirror duality, we obtain the dual accelerated mirror descent (dual-AMD) method that efficiently reduces $ψ^*(\nabla f(x))$, where $ψ$ is a distance-generating function and $ψ^*$ quantifies the magnitude of $\nabla f(x)$. We then apply dual-AMD to efficiently reduce $\|\nabla f(\cdot) \|_q$ for $q\in [2,\infty)$ and to efficiently compute $\varepsilon$-approximate solutions of the optimal transport problem.
title Mirror Duality in Convex Optimization
topic Optimization and Control
url https://arxiv.org/abs/2311.17296