Frank-Wolfe with Moreau Envelope Smoothing for Nonsmooth Nonconvex Problems
Antonio Silveti-Falls, Cesare Molinari, Zev Woodstock · 2026
Read Frank-Wolfe with Moreau Envelope Smoothing for Nonsmooth Nonconvex Problems by Antonio Silveti-Falls, Cesare Molinari, Zev Woodstock (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-00853.txt.
Description
Silveti-Falls, Molinari, and Woodstock smooth a hard class of optimization in this 2026 paper. Their target is nonsmooth, nonconvex problems under constraints. Frank-Wolfe methods suit such tasks but stall on kinks in the objective. Penalties like MCP and SCAD, and split tasks with many convex sets, fall in this class.
Their method, FRAMES, pairs each step with a smoothing pass. The Moreau envelope irons the objective's kinks, then one Frank-Wolfe step follows. They prove convergence, with rates, for both task classes under mild conditions. Even inconsistent problems, with no perfect point, are covered. A fresh link ties the Frank-Wolfe gap of the raw problem to that of its smoothed twin. That link shows past analyses were loose, they argue. Tests on matrix factorization and quadratic splitting bear the gains out in practice.
Catalog record
| CREATOR | Antonio Silveti-Falls, Cesare Molinari, Zev Woodstock |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 106 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.00853