Paper Image

Splitting Frank-Wolfe for intersection constraints

Published on:

9 November 2023

Primary Category:

Optimization and Control

Paper Authors:

Zev Woodstock,

Sebastian Pokutta

Bullets

Key Details

Proposes new conditional gradient algorithm for intersection constraints using individual linear minimization oracles

First convergence guarantees in nonconvex setting, with O(log(t)/sqrt(t)) rate

Combines product space reformulation with penalty method

Requires only one linear minimization oracle call per set per iteration

Connects sequence of relaxations to original problem via analytical results

AI generated summary

Splitting Frank-Wolfe for intersection constraints

This paper proposes a novel algorithm to efficiently solve optimization problems with intersection constraints, using linear minimization oracles for the individual sets. It is the first with convergence guarantees in the nonconvex setting. The approach combines a product space reformulation with a penalty method, requiring only one linear minimization oracle call per set per iteration.

Answers from this paper

Comments

No comments yet, be the first to start the conversation...

Sign up to comment on this paper

Sign Up