Edinburgh Research Explorer

Mini-Batch Primal and Dual Methods for SVMs

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Related Edinburgh Organisations

Original languageEnglish
Title of host publicationJMLR Workshop and Conference Proceedings
Subtitle of host publicationProceedings of the 30th International Conference on Machine Learning
Pages1022-1030
Volume28
Edition3
StatePublished - 2013

Abstract

We address the issue of using mini-batches in stochastic optimization of SVMs. We show that the same quantity, the spectral norm of the data, controls the parallelization speedup obtained for both primal stochastic subgradient descent (SGD) and stochastic dual coordinate ascent (SCDA) methods and use it to derive novel variants of mini-batched SDCA. Our guarantees for both methods are expressed in terms of the original nonsmooth primal problem based on the hinge-loss.

ID: 14690185