一类新型快速模糊支持向量机

A Kind of Novel Fast Fuzzy Support Vector Machines

针对一般模糊支持向量机训练时间过长,训练效率低下的问题,通过定义了一种新的隶属度函数的方法,来改进算法,从而得到了一种快速模糊支持向量机。本算法中的新定义的隶属度函数能够对离分类超平面较远、不可能成为支持向量的数据赋予较小的隶属度,使训练样本集中的数据大大减少。同时,在将二类模糊支持向量机推广到k类时,采用了DAGSVMs方法,进一步提高了多类分类问题的分类效率。实验表明,提出的快速模糊支持向量机在保证测试精度的同时,减少了训练时间。

Proposes a kind of fast fuzzy support vector machines to solve the problem of long training time and low training efficiency by improved algorithm. The definition of rnemhership function in the new algorithm can give the smaller memberships to the data away from separating hyperplane which can not be the support vector, so it reduces the data in the training sample set. Meanwhile, it expends to multi - classes fast fuzzy support vector machines using DAG. Experimental results indicate that the algorithm reduces the training time on the premise of guaranteeing the testing accuracy.