Skip to main content

Parallelizing Support Vector Machines


EMSL Project ID
35795

Abstract

The proposed project will focus on distributed and shared memory techniques to improve the execution efficiency and accuracy of Support Vector Machines (SVM). Current research suggests that SVM will play a role in classifying genomes and proteomes [1-4] and other fields of interest to EMSL, PNNL and DOE however, these techniques are computationally expensive, which limits their usefulness for large-scale problems such as community-level proteomic analysis. To address this problem, we are investigating parallelization techniques to take advantage of distributed and shared-memory computers.

Project Details

Start Date
2009-06-19
End Date
2010-06-20
Status
Closed

Team

Principal Investigator

Kevin Glass
Institution
Environmental Molecular Sciences Laboratory

Team Members

Patrick Nichols
Institution
Pacific Northwest National Laboratory

Christopher Oehmen
Institution
Pacific Northwest National Laboratory