基于边界向量的支持向量机增量算法

提出了一种新的基于边界向量的增量式支持向量机学习算法。该算法根据支持向量的几何分布特点,采用边界向量预选取方法,从增量样本中选取最有可能成为支持向量的样本形成边界向量集,在其上进行支持向量训练。通过对初始样本是否满足新增样本KKT条件的判断,解决非支持向量向支持向量的转化问题,有效地处理历史数据。针对UCI标准数据集上的仿真实验表明,基于边界向量的增量算法可以有效地减少训练样本数,积累历史信息,具有更高的分类速度和更好的推广能力。

Incremental learning algorithm of Support Vector Machine based on boundary vectors

A new incremental learning algorithm for the support vector machine based on boundary vector is proposed.The geometric character of the support vector is used to extract the simples,which can be the support vectors possibly from incremental simples.The method of the boundary vector pre-exacting is used to generate the boundary support vector set.It makes use of the condition of KKT to solve the problem that non-supports vectors transform to support vectors.And it can deal with the history simples efficiently.The experiment have been made on university of California-Irvine database.The result shows that the novel algorithm can greatly reduce the number of training data in incremental SVM training,accumulate the history information speed up the training process.