On a Frank-Wolfe Approach for Abs-smooth Functions
From MaRDI portal
Abstract: We propose an algorithm which appears to be the first bridge between the fields of conditional gradient methods and abs-smooth optimization. Our nonsmooth nonconvex problem setting is motivated by machine learning, since the broad class of abs-smooth functions includes, for instance, the squared -error of a neural network with ReLU or hinge loss activation. To overcome the nonsmoothness in our problem, we propose a generalization to the traditional Frank-Wolfe gap and prove that first-order minimality is achieved when it vanishes. We derive a convergence rate for our algorithm which is {em identical} to the smooth case. Although our algorithm necessitates the solution of a subproblem which is more challenging than the smooth case, we provide an efficient numerical method for its partial solution, and we identify several applications where our approach fully solves the subproblem. Numerical and theoretical convergence is demonstrated, yielding several conjectures.
This page was built for publication: On a Frank-Wolfe Approach for Abs-smooth Functions
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6429924)
