DocumentCode :
2379369
Title :
A novel speed-up SVM algorithm for massive classification tasks
Author :
Do, Thanh-Nghi ; Nguyen, Van-Hoa
Author_Institution :
Coll. of Inf. Technol., Can Tho Univ., Can Tho
fYear :
2008
fDate :
13-17 July 2008
Firstpage :
215
Lastpage :
220
Abstract :
The new parallel incremental support vector machine (SVM) algorithm aims at classifying very large datasets on graphics processing units (GPUs). SVM and kernel related methods have shown to build accurate models but the learning task usually needs a quadratic program so that the learning task for large datasets requires large memory capacity and long time. We extend a recent Least Squares SVM (LS-SVM) proposed by Suykens and Vandewalle for building incremental, parallel algorithm. The new algorithm uses graphics processors to gain high performance at low cost. Numerical test results on UCI, Delve dataset repositories showed that our parallel incremental algorithm using GPUs is about 65 times faster than a CPU implementation and often significantly over 1000 times faster than state-of-the-art algorithms LibSVM, SVM-perf and CB-SVM.
Keywords :
learning (artificial intelligence); least squares approximations; support vector machines; graphics processing units; kernel related methods; large memory capacity; massive classification tasks; quadratic program; speed-up SVM algorithm; support vector machine; very large datasets; Classification algorithms; Costs; Graphics; Kernel; Least squares methods; Parallel algorithms; Performance gain; Support vector machine classification; Support vector machines; Testing; data mining; graphics processing unit; incremental learning; least squares support vector machine; machine learning; massive data classification; parallel algorithm;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Research, Innovation and Vision for the Future, 2008. RIVF 2008. IEEE International Conference on
Conference_Location :
Ho Chi Minh City
Print_ISBN :
978-1-4244-2379-8
Electronic_ISBN :
978-1-4244-2380-4
Type :
conf
DOI :
10.1109/RIVF.2008.4586358
Filename :
4586358
Link To Document :
بازگشت