Asynchronous Parallel Block-Coordinate Frank-Wolfe
Year: 2014
Type: preprint
Abstract: We develop mini-batched parallel Frank-Wolfe (conditional gradient) methods for smooth convex optimization subject to block-separable constraints. Our work includes the basic (batch) Frank-Wolfe algorithm as well as the recently proposed Block-Coordinate Frank-Wolfe (BCFW) method\citep{lacoste2012block} as special cases. Our algorithm permits asynchronous updates within the minibatch, and is robust to stragglers and faulty worker threads. Our analysis reveals how the potential speedups over BCFW depend on the minibatch size and how one can provably obtain large problem dependent speedups. We present several experiments to indicate empirical behavior of our methods, obtaining significant speedups over competing state-of-the-art (and synchronous) methods on structural SVMs.
Source: arXiv (Cornell University)
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2160570933