摘要： 针对多类分类问题中现有算法精度不高的问题，基于一类分类马氏椭球学习机，提出一种最大间隔椭球形多类分类算法，将每一类数据用超椭球来界定，数据空间由若干个超椭球组成，每个超椭球包围一类样本点，并以最大间隔排除不属于该类的样本点，该算法同时考虑了不同类样本点的协方差矩阵，即分布信息。真实数据上的实验结果表明该方法能提高分类精度。

关键词: 模式识别, 多类分类, 最大间隔, 超椭球

Abstract: For the problem of low accuracy in existing multi-class classification algorithm, based on Mahalanobis ellipsoidal learning machine for one class classification, a maximal margin ellipsoid-shaped multi-class classification algorithm is proposed, which bounds each class data using a hyper-ellipsoid and the data space is composed of several hyper-ellipsoids. Each hyper-ellipsoid encloses all samples from one class and at the same time excludes all samples from the rest class with maximal margin. In addition, the covariance matrix, i.e., the distribution information of examples from different classes is considered. Experimental results on real data sets show that the method can improve accuracy for classification.

Key words: pattern recognition, multi-class classification, maximal margin, hyper-ellipsoid

中图分类号: 

• TP181