支撑矢量机的改进分类器算法

Modifying classifier method for support vector machine

支撑矢量机是 2 0世纪 90年代中期发展起来的机器学习技术 ,改进分类器算法通过增大广义最优超平面的分类间隔 ,实现了识别能力的提高。在此基础之上 ,预选取部分训练样本 ,来提高优化速度 ,而且不会降低分类能力 ,从而能够同时提高支撑矢量机的识别率和降低时间复杂度 ,为支撑矢量机的应用提供了一种有效的实用化方法。实验结果表明 ,该方法在可分性得到显著提高的同时提高了速度。

Support vector machine is a new machine learning technique developed from the middle of 1990s.Being different from traditional neural networks,it is based on structure risk minimization principle.In this paper,modifying classifier method is proposed,which can improve the performance of a support vector machine classifier by a conformal mapping.But the modifying classifier method is accomplished by two optimization.In order to improve the speed of support vector machine,the training patterns are reduced through the pre extracting support vector machine.The experimental results show that the separability between classes is increased and speed is well improved.