摘要
提出了一种支持向量预提取方法,对核感知机误分次数与决策边界的关系作了分析,引入了核函数和误分界,利用结构简单的感知机算法构建支持向量预提取模块,压缩样本规模,然后将处理结果输入支持向量机进行再处理,在精度和处理速度方面取得了较好的效果,仿真实验验证了这一方法的可行性。
This Paper analyzed the shortcomings of Support Vector Machine (SVM) dealing with large sample sets,proposed a support vector pre-extraction method. Firstly,used perceptron algorithm to extract semi-support vectors based on the relationship between error times and decision border. Secondly, input them into SVM. It can effectively utilize the simpleness of perceptron algorithm and excellent performance of SVM on linear and nonlinear questions. Experiment results show that this method seems to have some more applied value ,especially on large sample sets in this domain.
出处
《广西师范大学学报(自然科学版)》
CAS
北大核心
2006年第4期123-126,共4页
Journal of Guangxi Normal University:Natural Science Edition
基金
国家自然科学基金资助项目(10576027)
四川省应用基础研究基金资助项目(05JY029-006-4)
关键词
信号处理
感知机
核感知支持向量机
支持向量预提取
近似支持向量
signal process
perceptron algorithm
kernel perceptron support vector machine (KP-SVM)
support vectors pre-extraction
semi-support vectors