基于自适应邻域选择的局部判别投影算法

Local Discriminant Projection Algorithm Based on Adaptive Neighborhood Selection

在多模数据分类中,使用局部Fisher判别分析和边界Fisher分析方法构建邻域不能充分反映流形学习对邻域的要求。为此,提出一种基于自适应邻域选择的局部判别投影算法。采用自适应方法扩大或者缩小近邻系数k,以构建邻域,从而保持局部线性结构,揭示流形的内在几何结构,利用局部化方法使得投影空间中同类近邻样本尽量紧凑、异类近邻样本尽量分开。在ORL和YALE人脸数据库中进行实验,结果表明,在不同训练样本个数下,该算法均能获得较高的识别率。

Aiming at the drawback that Local Fisher Discriminant Analysis（LFDA）algorithm and the Marginal Fisher Analysis（MFA） algorithm solve the problem of multimodal data classification and construct a reasonable neighborhood for each point. A local discriminant projection algorithm based on adaptive neighborhood selection is proposed in this paper. An adaptive algorithm to expand or narrow neighbor coefficient k is adopted to keep the local linear structure. So it perfectly detects the intrinsic geometric structure of manifold. The underlying idea of the new method is that the desired projection should make neighbors of the same class close and neighbors of different classes apart. Doing test on the ORL and the YALE face database, the results show that this algorithm can achieve higher recognition rate under different training samples.