Blended Conditional Gradients: the unconditioning of conditional gradients

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Braun, Gábor, Pokutta, Sebastian, Tu, Dan, Wright, Stephen
Format: Preprint
Published: 2018
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912284315484160
author Braun, Gábor
Pokutta, Sebastian
Tu, Dan
Wright, Stephen
author_facet Braun, Gábor
Pokutta, Sebastian
Tu, Dan
Wright, Stephen
contents We present a blended conditional gradient approach for minimizing a smooth convex function over a polytope P, combining the Frank--Wolfe algorithm (also called conditional gradient) with gradient-based steps, different from away steps and pairwise steps, but still achieving linear convergence for strongly convex functions, along with good practical performance. Our approach retains all favorable properties of conditional gradient algorithms, notably avoidance of projections onto P and maintenance of iterates as sparse convex combinations of a limited number of extreme points of P. The algorithm is lazy, making use of inexpensive inexact solutions of the linear programming subproblem that characterizes the conditional gradient approach. It decreases measures of optimality (primal and dual gaps) rapidly, both in the number of iterations and in wall-clock time, outperforming even the lazy conditional gradient algorithms of [arXiv:1410.8816]. We also present a streamlined version of the algorithm for the probability simplex.
format Preprint
id arxiv_https___arxiv_org_abs_1805_07311
institution arXiv
publishDate 2018
record_format arxiv
spellingShingle Blended Conditional Gradients: the unconditioning of conditional gradients
Braun, Gábor
Pokutta, Sebastian
Tu, Dan
Wright, Stephen
Optimization and Control
Computational Complexity
Machine Learning
68Q32, 90C52
We present a blended conditional gradient approach for minimizing a smooth convex function over a polytope P, combining the Frank--Wolfe algorithm (also called conditional gradient) with gradient-based steps, different from away steps and pairwise steps, but still achieving linear convergence for strongly convex functions, along with good practical performance. Our approach retains all favorable properties of conditional gradient algorithms, notably avoidance of projections onto P and maintenance of iterates as sparse convex combinations of a limited number of extreme points of P. The algorithm is lazy, making use of inexpensive inexact solutions of the linear programming subproblem that characterizes the conditional gradient approach. It decreases measures of optimality (primal and dual gaps) rapidly, both in the number of iterations and in wall-clock time, outperforming even the lazy conditional gradient algorithms of [arXiv:1410.8816]. We also present a streamlined version of the algorithm for the probability simplex.
title Blended Conditional Gradients: the unconditioning of conditional gradients
topic Optimization and Control
Computational Complexity
Machine Learning
68Q32, 90C52
url https://arxiv.org/abs/1805.07311