Abstract: This paper proposes a locally adaptive nearest neighbor classification method based on supervised learning style which works well for the classes more than two. In this method, the ellipsoid clustering learning method is applied to estimate an effective metric for producing neighborhood that is elongated along less discriminating feature dimensions and constricted along most discriminating ones. As a result, the class conditional probabilities can be expected to be approximately constant in the modified neighborhoods, whereby better classification performance can be achieved. The experimental results show that this is an efficient and robust classification method for multi-class problems.

Key words: supervised ellipsoid clustering, nearest neighbor classifiers, multi-class

摘要： 基于多类别监督学习，提出了一种局部自适应最近邻分类器。此方法使用椭球聚类学习方法估计有效尺度，用于拉长特征不明显的维，并限制特征重要的维。在修正的领域中，类条件概率按预期近似为常数，从而得到更好的分类性能。实验结果显示，对多类问题，这是一种有效且鲁棒的分类方法。

关键词: 监督椭球聚类学习, 最近邻分类器, 多类

CLC Number: 

• TP311