The Iterates of the Frank-Wolfe Algorithm May Not Converge
Abstract
The Frank-Wolfe algorithm is a popular method for minimizing a smooth convex function over a compact convex set . While many convergence results have been derived in terms of function values, hardly nothing is known about the convergence behavior of the sequence of iterates . Under the usual assumptions, we design several counterexamples to the convergence of , where is -time continuously differentiable, , and . Our counterexamples cover the cases of open-loop, closed-loop, and line-search step-size strategies. We do not assume \emph{misspecification} of the linear minimization oracle and our results thus hold regardless of the points it returns, demonstrating the fundamental pathologies in the convergence behavior of .
Cite
@article{arxiv.2202.08711,
title = {The Iterates of the Frank-Wolfe Algorithm May Not Converge},
author = {Jérôme Bolte and Cyrille W. Combettes and Édouard Pauwels},
journal= {arXiv preprint arXiv:2202.08711},
year = {2022}
}
Comments
15 pages, 7 figures