Saved in:
Bibliographic Details
Main Authors: Mehta, Ronak, Diakonikolas, Jelena, Harchaoui, Zaid
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2403.10763
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929708177817600
author Mehta, Ronak
Diakonikolas, Jelena
Harchaoui, Zaid
author_facet Mehta, Ronak
Diakonikolas, Jelena
Harchaoui, Zaid
contents We consider the penalized distributionally robust optimization (DRO) problem with a closed, convex uncertainty set, a setting that encompasses learning using $f$-DRO and spectral/$L$-risk minimization. We present Drago, a stochastic primal-dual algorithm that combines cyclic and randomized components with a carefully regularized primal update to achieve dual variance reduction. Owing to its design, Drago enjoys a state-of-the-art linear convergence rate on strongly convex-strongly concave DRO problems with a fine-grained dependency on primal and dual condition numbers. Theoretical results are supported by numerical benchmarks on regression and classification tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2403_10763
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Drago: Primal-Dual Coupled Variance Reduction for Faster Distributionally Robust Optimization
Mehta, Ronak
Diakonikolas, Jelena
Harchaoui, Zaid
Machine Learning
Optimization and Control
We consider the penalized distributionally robust optimization (DRO) problem with a closed, convex uncertainty set, a setting that encompasses learning using $f$-DRO and spectral/$L$-risk minimization. We present Drago, a stochastic primal-dual algorithm that combines cyclic and randomized components with a carefully regularized primal update to achieve dual variance reduction. Owing to its design, Drago enjoys a state-of-the-art linear convergence rate on strongly convex-strongly concave DRO problems with a fine-grained dependency on primal and dual condition numbers. Theoretical results are supported by numerical benchmarks on regression and classification tasks.
title Drago: Primal-Dual Coupled Variance Reduction for Faster Distributionally Robust Optimization
topic Machine Learning
Optimization and Control
url https://arxiv.org/abs/2403.10763