Principal component analysis (PCA) is a kind of efficient method for feature extraction and compression, but the heavy burden of computation and storage restrict its realistic application, its incremental algorithm ca...
详细信息
Principal component analysis (PCA) is a kind of efficient method for feature extraction and compression, but the heavy burden of computation and storage restrict its realistic application, its incremental algorithm can resolve above question efficiently, in this paper CCIPCA is investigated. Through the theory analysis, the CCIPCA can be seen as a kind of online form of power method, and its statistical efficiency corresponds to the situation when the data is weighted with exponential window. and the parameter , Based on the character of power method, the initial vector and weight were adjusted, and the computation burden of CCIPCA was decreased remarkably, Based on the modified algorithm, an incremental BDPCA algorithm was provided too. Experiments based on ORL and ATT face database testify the efficiency of the method.
暂无评论