A Fast Method for Training Linear SVM in the Primal

Trinh Minh Tri Do 1 Thierry Artières 1
1 MALIRE - Machine Learning and Information Retrieval
LIP6 - Laboratoire d'Informatique de Paris 6
Abstract : We propose a new algorithm for training a linear Support Vector Machine in the primal. The algorithm mixes ideas from non smooth optimization, subgradient methods, and cutting planes methods. This yields a fast algorithm that compares well to state of the art algorithms. It is proved to require O(1/λε) iterations to converge to a solution with accuracy ε. Additionally we provide an exact shrinking method in the primal that allows reducing the complexity of an iteration to much less than O(N) where N is the number of training samples.
Document type :
Conference papers
Complete list of metadatas

https://hal.archives-ouvertes.fr/hal-01301600
Contributor : Lip6 Publications <>
Submitted on : Tuesday, April 12, 2016 - 3:05:30 PM
Last modification on : Thursday, September 19, 2019 - 2:20:04 PM

Links full text

Identifiers

Citation

Trinh Minh Tri Do, Thierry Artières. A Fast Method for Training Linear SVM in the Primal. European Conference on Machine Learning (ECML), Sep 2008, Antwerp, Belgium. pp.272-287, ⟨10.1007/978-3-540-87479-9_36⟩. ⟨hal-01301600⟩

Share

Metrics

Record views

137