中文摘要：

提出了利用遗传算法对多核支持向量机的权系数进行寻优的方法GA-MKL,先选择表现能力最好的参数的单核构成多核,再利用遗传算法来对多核的核系数来寻优.采用该算法在UCI标准数据集上进行了实验,结果表明,该算法为多核SVM的系数选择提供了一种可行的方法.与单核SVM相比,该方法具有更好的分类能力,和其他多核学习算法相比,性能也有一定的提高.

英文摘要：

In this paper we present a method, GA-MKL, by using the genetic algorithm to optimize the weights of linear multi-kernel support vector machine. We first implement single-kernel experiment to decide the best parameter of each kernel then construct the linear multi-kernel with these indeterminate weights. Then we utilize the genetic al- gorithm to find the best weights which give the best classification performance. The classification experiments on the UCI database are employed with this algorithm. By comparison with the single-kernel algorithm, the experimental results show that this algorithm is superior to the single-kernel, thus it provides a feasible method for finding the best weights for multi-kernel SVM. This algorithm performs better than other MKL algorithm as well.