基于监督学习的核拉普拉斯特征映射分类方法

Classification based on supervised kernel Laplacian eigienmaps

提出一种监督学习的核拉普拉斯特征映射方法(supervised kernel Laplacian eigenmap,SKLE),通过非线性核映射将样本数据投影到高维核特征空间,然后将流形结构和样本类别信息进行有效的结合后,提取嵌入在高维数据中的低维流形特征用于分类.实验表明,该方法对新样本具有泛化性,并且能有效提高分类的效能.

Proposes a method named supervised kernel Laplacian eigenmaps （SKI.E）, which suggests using the kernel non - linear mapping to project the sample data onto the high - dimensional kernel characteristic space, and then combining the samples of manifold architecture and category information effectively, and finally extracting the low - dimensional manifolds features embedded in high - dimensional data for classification. Experiments show that the method has a generalization performance to new samples, and can effectively improve the classification performance.