摘要
针对传统SVM对噪声点和孤立点敏感的问题,以及不能解决样本特征规模大、含有异构信息、在特征空间中分布不平坦的问题,将模糊隶属度融入多核学习中,提出了一种模糊多核学习的方法;通过实验验证了模糊多核学习比传统SVM、模糊支持向量机以及多核学习具有更好的分类效果,从而验证了所提方法能够有效的克服传统SVM对噪声点敏感以及数据分布不平坦的问题。
According to issues on noise and isolated points sensitive,the traditional SVM can not solve the sample characteristics of large scale,containing the heterogeneous information and the uneven feature space distribution.The fuzzy membership is draw into multiple kernel learning,and this paper proposes a method of fuzzy multiple kernel learning.Experimental results show that fuzzy multiple kernel learning is more effective and feasible than SVM,Fuzzy support vector machine and Multiple kernel learning.So,proposed method in this paper can effectively overcome the traditional SVM on noise sensitive and data distribution uneven problem.
出处
《计算机测量与控制》
2016年第3期231-233,共3页
Computer Measurement &Control
基金
重庆市自然科学项目(cstc2014jcyjA40011)
重庆市教委科技项目(KJ1400513)
关键词
支持向量机
模糊支持向量机
多核学习
模糊多核学习
support vector machine
fuzzy support vector machine
multiple kernel learning
fuzzy multiple kernel learning