Block-Coordinate Frank-Wolfe Optimization for Structural SVMs
Abstract
We propose a randomized block-coordinate variant of the classic Frank-Wolfe algorithm for convex optimization with block-separable constraints. Despite its lower iteration cost, we show that it achieves a similar convergence rate in duality gap as the full Frank-Wolfe algorithm. We also show that, when applied to the dual structural support vector machine (SVM) objective, this yields an online algorithm that has the same low iteration complexity as primal stochastic subgradient methods. However, unlike stochastic subgradient methods, the block-coordinate Frank-Wolfe algorithm allows us to compute the optimal step-size and yields a computable duality gap guarantee. Our experiments indicate that this simple algorithm outperforms competing structural SVM solvers.
Cite
Text
Lacoste-Julien et al. "Block-Coordinate Frank-Wolfe Optimization for Structural SVMs." International Conference on Machine Learning, 2013.Markdown
[Lacoste-Julien et al. "Block-Coordinate Frank-Wolfe Optimization for Structural SVMs." International Conference on Machine Learning, 2013.](https://mlanthology.org/icml/2013/lacostejulien2013icml-blockcoordinate/)BibTeX
@inproceedings{lacostejulien2013icml-blockcoordinate,
title = {{Block-Coordinate Frank-Wolfe Optimization for Structural SVMs}},
author = {Lacoste-Julien, Simon and Jaggi, Martin and Schmidt, Mark and Pletscher, Patrick},
booktitle = {International Conference on Machine Learning},
year = {2013},
pages = {53-61},
volume = {28},
url = {https://mlanthology.org/icml/2013/lacostejulien2013icml-blockcoordinate/}
}