Project
EF1-23
On a Frank-Wolfe Approach for Abs-Smooth Optimization
Project Heads
Sebastian Pokutta, Andrea Walther, Zev Woodstock (until 08/25)
Project Members
Sri Harshitha Tadinada
Project Duration
01.04.2023 − 30.09.2026
Located at
HU Berlin
Description
Motivated by nonsmooth problems in machine learning, we solve the problem of minimizing an abs-smooth function subject to closed convex constraints. New theory and algorithms are developed using linear minimization oracles to enforce constraints and abs-linearization methods to handle nonsmoothness.
Related Publications
Timo Kreimeier, Sebastian Pokutta, Andrea Walther, and Zev Woodstock. On a Frank-Wolfe Approach for Abs-smooth Functions.
Sri Harshitha Tadinada, Tim Siebert, Jürgen Fuhrmann, Sebastian Pokutta and Andrea Walther. An AD-enabled Frank-Wolfe method for non-smooth optimization.
Sri Harshitha Tadinada, Sebastian Pokutta and Andrea Walther. Abs-Smooth Frank-Wolfe Method: Primal-Dual Analysis, Heavy Ball Momentum, and Inexact Oracles.
Related Software