SwePub - search: WFRF:(Rätsch Gunnar)

Enumeration	Reference	Cover	Find
1.	Dresdner, Gideon, et al. (author) Faster One-Sample Stochastic Conditional Gradient Method for Composite Convex Minimization 2022 In: Proceedings of The 25th International Conference on Artificial Intelligence and Statistics. - : PMLR. Conference paper (peer-reviewed)abstract We propose a stochastic conditional gradient method (CGM) for minimizing convex finite-sum objectives formed as a sum of smooth and non-smooth terms. Existing CGM variants for this template either suffer from slow convergence rates, or require carefully increasing the batch size over the course of the algorithm's execution, which leads to computing full gradients. In contrast, the proposed method, equipped with a stochastic average gradient (SAG) estimator, requires only one sample per iteration. Nevertheless, it guarantees fast convergence rates on par with more sophisticated variance reduction techniques. In applications we put special emphasis on problems with a large number of separable constraints. Such problems are prevalent among semidefinite programming (SDP) formulations arising in machine learning and theoretical computer science. We provide numerical experiments on matrix completion, unsupervised clustering, and sparsest-cut SDPs.

Dresdner, Gideon, et al. (author)
Faster One-Sample Stochastic Conditional Gradient Method for Composite Convex Minimization
2022
In: Proceedings of The 25th International Conference on Artificial Intelligence and Statistics. - : PMLR.
Conference paper (peer-reviewed)abstract
- We propose a stochastic conditional gradient method (CGM) for minimizing convex finite-sum objectives formed as a sum of smooth and non-smooth terms. Existing CGM variants for this template either suffer from slow convergence rates, or require carefully increasing the batch size over the course of the algorithm's execution, which leads to computing full gradients. In contrast, the proposed method, equipped with a stochastic average gradient (SAG) estimator, requires only one sample per iteration. Nevertheless, it guarantees fast convergence rates on par with more sophisticated variance reduction techniques. In applications we put special emphasis on problems with a large number of separable constraints. Such problems are prevalent among semidefinite programming (SDP) formulations arising in machine learning and theoretical computer science. We provide numerical experiments on matrix completion, unsupervised clustering, and sparsest-cut SDPs.

Träfflista för sökning "WFRF:(Rätsch Gunnar) srt2:(2022)"