基于图嵌入的特征提取算法研究

通过对图、完全图和正则图概念的介绍,详细地描述了图嵌入的方法,同时对主成分分析、线性鉴别分析、局部保持投影、保持近邻嵌入、L1图及其嵌入等经典的特征提取算法进行了详细的代数推导,列出了详细的推导过程,得出这些经典算法可以用图嵌入理论来解释的结论,最后得出特征提取算法的核心在于算法的图构造.

Research on Extraction Algorithm Based on Graph Embedding Feature

By introducing graph, complete graph and regular graph conception, we elaborate the definition of graph embedding. Mean-while, this paper derives algebraically those classic feature extraction algorithm including Principal Component Analysis, linear discriminant analysis, locality preserving projections, neighborhood preserving embedding, L1 graph and embedding, etc. Derivation process is described in detail and a conclusion, which is all of these classic algorithms can be explained by graph embedding theory, is drawn. In conclusion, the core of feature extraction algorithm is graph construction of algorithm.