分类大规模数据的核向量机方法研究

Study on the core vector machine method for the classification of large-scale data

标准的支持向量机算法需要求解二次规划问题,因此,在处理大规模样本的时候,求解二次规划问题的时间复杂度和空间复杂度就成为支持向量机应用的一个瓶颈.核向量机将传统支持向量机中的二次规划问题转化为求解最小包围球问题,从而显著降低了二次规划的复杂程度.使用核向量机对大规模数据进行分类,所选用的数据样本数均超过2000,并与标准的支持向量机作了对比实验结果表明:核向量机在处理大规模数据分类时,比标准的支持向量机计算复杂度低,训练速度快,耗费空间少.

The standard support vector machine algorithm needs to solve the quadratic programming problem.Therefore,when large-scale samples are processed,the time complexity and space complexity of solving the quadratic programming problem become a bottleneck in the applications of the support vector machine.Core vector machine converts the solution of the quadratic programming problem in the traditional support vector machine algorithm into the problem of solving minimum enclosed ball,which greatly reduces the complexity of the quadratic programming problem.In this paper,the core vector machine is used for the classification of the large-scale samples,and the number of these samples is over 2 000.The classification process of this method is compared with that of the standard support vector machine method,and the results show that the core vector machine method has a lower computational complexity,higher training speed and less space cost.