Dr Jussi Kujala (Tampere University of Téchnology) will give a talk on
Linear SVMs and 1-norm problems
on Wednesday 6 May at 12:15 in C221, Kumpula.
Welcome!
Heikki
---
Abstract
Recently the machine learning community has given birth to several linear SVM solvers, such as Pegasos, SGD, OCAS, Liblinear, and svmperf. These are scalable in contrast to their more general counterparts such as svmlight. We briefly describe some implications of these algorithms in solving something similar to 1-norm SVM, which is an SVM but with regularization that resists overfitting to irrelevant attributes. Also, we recap connections between linear SVM and SVM.
Last updated on 6 May 2009 by Visa Noronen - Page created on 6 May 2009 by Visa Noronen