Title :
SVM parameters tuning with quantum particles swarm optimization
Author :
Luo, Zhiyong ; Zhang, Wenfeng ; Li, Yuxia ; Xiang, Min
Author_Institution :
Sch. of Autom., Chongqing Univ. of Posts & Telecommun., Chongqing
Abstract :
Common used parameters selection method for support vector machines (SVM) is cross-validation, which is complicated calculation and takes a very long time. In this paper, a novel regularization parameter and kernel parameter tuning approach of SVM is presented based on quantum particle swarm optimization algorithm (QPSO). QPSO is a particle swarm optimization (PSO) with quantum individual that has better global search capacity. The parameters of least squares support vector machines (LS-SVM) can be adjusted using QPSO. Classification and function estimation are studied using LS-SVM with wavelet kernel and Gaussian kernel. The simulation results show that the proposed approach can effectively tune the parameters of LS-SVM, and improved LS-SVM with wavelet kernel can provide better precision.
Keywords :
Gaussian processes; least squares approximations; particle swarm optimisation; pattern classification; quantum computing; search problems; support vector machines; wavelet transforms; Gaussian kernel; SVM parameters tuning; classification estimation; function estimation; global search capacity; least squares support vector machines; quantum particles swarm optimization; wavelet kernel; Automation; Kernel; Learning systems; Least squares approximation; Least squares methods; Particle swarm optimization; Pattern recognition; Risk management; Support vector machine classification; Support vector machines; least squares support vector machines (LS-SVM); parameters tuning; quantum particle swarm optimization algorithm (QPSO); support vector machines (SVM);
Conference_Titel :
Cybernetics and Intelligent Systems, 2008 IEEE Conference on
Conference_Location :
Chengdu
Print_ISBN :
978-1-4244-1673-8
Electronic_ISBN :
978-1-4244-1674-5
DOI :
10.1109/ICCIS.2008.4670970