339. Revisiting Frank-Wolfe for Nonconvex Problems
Invited abstract in session FD-6: Difference and decomposition methods, stream Methods for non-/monotone inclusions and their applications.
Friday, 14:10 - 15:50Room: M:H
Authors (first author is the speaker)
| 1. | Hoomaan Maskan
|
| Umeå University | |
| 2. | Suvrit Sra
|
| Massachusetts Institute of Technology | |
| 3. | Alp Yurtsever
|
| Umeå University |
Abstract
Difference of convex (DC) programs are a class of nonconvex problems that address a wide range of applications. Therefore, there has been a growing effort to solve these problems more efficiently. In this talk, we introduce a Frank-Wolfe algorithm for minimizing difference of convex functions over a convex and compact set. Our analysis indicates that the proposed method converges to a stationary point of the problem at an O(1/sqrt(k)) rate. Notably, while this rate resembles that of nonconvex Frank-Wolfe methods, numerical simulations demonstrate that our proposed approach often identifies a better local optimal point with a smaller objective value compared to the standard nonconvex Frank-Wolfe algorithm.
Keywords
- Global optimization
Status: accepted
Back to the list of papers