一种模糊支持向量机主动学习算法  

AN ACTIVE LEARNING ALGORITHM OF FUZZY SUPPORT VECTOR MACHINE

作　　者：孙秀英[1] 王燕[1] 

机构地区：[1]黄河科技学院现代教育技术中心,河南郑州450006

出　　处：《计算机应用与软件》2013年第10期136-138,共3页Computer Applications and Software

基　　金：河南省教育厅科学技术研究重点项目(12B880010)

摘　　要：模糊支持向量机具有很好的抗噪声能力,受到很多专家的重视。然而模糊支持向量机上的主动学习算法却一直鲜有研究。提出一种针对模糊支持向量机的主动学习算法,该算法首先在训练集合上利用模糊支持向量机得到决策超平面,然后选取间隔内的未标记样本进行标记,并计算相应的模糊权重,以及更新原有训练样本的模糊权重。最后结合赋予模糊权重的新标记样本重新进行学习,直到未标记样本集为空或者分类性能满足要求。在UCI标准数据集和语音识别数据集上的实验充分验证了该算法的有效性。Fuzzy support vector machine （FSVM） has good anti-noise performance, which attracts the attention from many experts. However, the active learning algorithm of FSVM has been seldom studied. In this paper, we propose an active learning algorithm for FSVM. This algorithm firstly obtains the decision-making hyperplane using FSVM from the training collection. Then it marks the unlabeled samples selected from the interval, and computes their corresponding fuzzy weights, and updates the fuzzy weights of original training samples. Finally, the algorithm makes re-leaming in combination with the new labelled samples assigned with fuzzy weights until the unlabeled samples set to be empty or the classification performance meeting the requirements. The experiments on UCI standard dataset and speech recognition dataset fully verify the effectiveness of the proposed algorithm.

关 键 词：模糊支持向量机 抗噪 主动学习 间隔 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]