论文部分内容阅读
针对支持向量机对训练样本中的噪声和孤立点特别敏感的问题,提出一种基于边界向量提取的模糊支持向量机方法.在特征空间中寻找能够分别包住两类样本点的两个最小超球,并选择可能成为支持向量的边界向量作为新样本,减少参与训练的样本数目,提高训练速度.样本的隶属度根据边界样本和噪声点与所在超球球心的距离分别确定,既减弱孤立点和噪声的影响,又增强支持向量对支持向量机分类的作用.实验结果表明,与传统的支持向量机方法和基于样本与类中心之间关系的模糊支持向量机相比,本文方法具有更快的学习速度和更好的泛化能力.
Aiming at the problem that SVM is particularly sensitive to noise and outliers in training samples, a fuzzy support vector machine method based on boundary vector extraction is proposed. In the feature space, two minimal super And select the boundary vector which may become the support vector as a new sample to reduce the number of samples participating in training and improve the training speed.The membership of the sample is determined according to the distance between the boundary sample and the noise point and the hypersphere, Point and noise, and enhances the support vector classification for SVM.The experimental results show that, compared with the traditional support vector machine method and the fuzzy support vector machine based on the relationship between the sample and the class center, this method has more Quick learning speed and better generalization ability.