University of Illinois Urbana-Champaign

Solving Very Large Scale Linear SVM Using Multiple Processors

Jindal, Prateek; Roth, Dan; Kale, Laxmikant V.

Permalink

https://hdl.handle.net/2142/46407

Description

Title
Solving Very Large Scale Linear SVM Using Multiple Processors
Author(s)
  • Jindal, Prateek
  • Roth, Dan
  • Kale, Laxmikant V.
Issue Date
2013-12-04
Keyword(s)
  • Support Vector Machines
  • Big Data Analytics
  • Charm++
Abstract
SVMs have been used for long for data classification. While solving very large problems, one may encounter hundreds of thousands of features and a large number of training vectors. A natural solution to solving problems with large datasets is to use multiple processors. In this report, we discuss the parallel implementation of Liblinear which is a very good sequential tool for using SVMs.
Type of Resource
text
Language
en
Permalink
http://hdl.handle.net/2142/46407
Copyright and License Information
2013 by Prateek Jindal

Owning Collections

Research and Tech Reports - Computer Science PRIMARY