Title :
Improved decomposition method for support vector machines
Author :
Zhou, Weida ; Zhang, Li ; Jiao, Licheng
Author_Institution :
National Key Lab. for Radar Signal Process., Xidian Univ., Xi´´an, China
Abstract :
An improved decomposition method for SVMs is presented. In our method, we propose a rule for selecting the working set to improve the training speed. The selection rule can be applied to any decomposition methods for SVMs. Simulation results show the feasibility and validity of our algorithm.
Keywords :
quadratic programming; support vector machines; very large databases; decomposition method; large databases; quadratic programming; selection rule; support vector machines; training patterns; training speed; Automation; Databases; Indexes; Iterative algorithms; Iterative methods; Large-scale systems; Quadratic programming; Radar signal processing; Signal processing algorithms; Support vector machines;
Conference_Titel :
Computational Intelligence and Multimedia Applications, 2003. ICCIMA 2003. Proceedings. Fifth International Conference on
Print_ISBN :
0-7695-1957-1
DOI :
10.1109/ICCIMA.2003.1238096
