Improvements to Platt's SMO algorithm for SVM classifier design

Authors
Keerthi, SS Shevade, SK Bhattacharyya, C Murthy, KRK
Citation
Ss. Keerthi et al., Improvements to Platt's SMO algorithm for SVM classifier design, NEURAL COMP, 13(3), 2001, pp. 637-649
Citations number
12
Categorie Soggetti
Neurosciences & Behavoir","AI Robotics and Automatic Control
Journal title
NEURAL COMPUTATION
ISSN journal
08997667 → ACNP
Volume
13
Issue
3
Year of publication
2001
Pages
637 - 649
Database
ISI
SICI code
0899-7667(200103)13:3<637:ITPSAF>2.0.ZU;2-9
Abstract
This article points out an important source of inefficiency in Platt's sequ ential minimal optimization (SMO) algorithm that is caused by the use of a single threshold value. Using clues from the KKT conditions for the dual pr oblem, two threshold parameters are employed to derive modifications of SMO . These modified algorithms perform significantly faster than the original SMO on all benchmark data sets tried.