关键词: 新增样本点延拓, 稀疏编码, 局部结构

Abstract: To solve out-of-sample problem of Laplacian Eigenmaps, a method preserving local structure is proposed which is based on the assumption that there is a linear relationship between the new sample and its neighbors. Then sparse-coding is used to obtain the linear reconstruction coefficients between the new sample and its neighbors. Finally, the low-dimensional representation of the new sample is computed through the linear relationship. The classification of the low-
dimensional representation is made by 1-NN classifier. Compared with sparse-coding reconstruction method based on global relationship, the method based on local information achieves higher accuracy using less time showing its superiority. Furthermore, the proposed method can be easily extended to the out-of-sample problem of other non-linear dimensionality reduction methods.

Key words: out-of-sample extension, sparse-coding, local structure