摘要
分析了支持向量机(support vector machine,SVM)目前主要存在的问题和参数选择对分类性能的影响后,提出了以改进粒子群算法优化SVM关键参数的优化SVM算法。将加入拥挤度因子的微粒群算法引入到SVM中,在不牺牲泛化性能的前提下,对其参数进行优化,增加了SVM初始化参数的多样性,减慢了局部搜索,促进其在全局范围内的寻优搜索,以有效克服SVM算法过分依赖初始值和容易陷入局部极小值的缺点,并利用由粗到精的策略构造多层SVM人脸表情分类器,在提高准确率的基础上加快分类的速度。实验证明,新算法具有速度快、准确率高的优点。
This paper analyzed current problems of SVM and the effect of parameter selection on classification performance. It presented an optimized SVM algorithm based on improving SVM key parameters by using improved particle swarm optimiza- tion. It introduced particle swarm algorithm with crowded factor into SVM. It optimized the parameters without sacrificing generalization ability, also increased the diversity of SVN initialization parameter. This method slowed down local search. On the other hand, it enhanced the optimization search ability within the global scope so as to overcome the SVN algorithm' s shortcomings of over-reliance on the initial value and easy to fall into the local minimum value. By using coarse to precious strategy, it built up multi-class SVM face expression classifier. It accelerated classification speed on the basis of increase the accuracy. Experiment results show the fast speed and high accuracy of the new algorithm.
出处
《计算机应用研究》
CSCD
北大核心
2013年第8期2541-2544,共4页
Application Research of Computers
基金
国家自然科学基金资助项目(60973095)
关键词
支持向量机
改进粒子群优化
人脸表情分类器
support vector machine(SVM)
improved particle swarm optimization(IPSO)
face expression classifier