An Accurate Incremental Principal Component Analysis method with capacity of update and downdate

摘要

Principal Component Analysis is a popular and powerful method in many machine learning task, The traditional PCA is implemented in batch mode, which means the much lower efficiency, especially for the task which training dataset is updated or downdated frequently, so it is reasonable to develop the incremental version of PCA. However, most of the existing incremental PCA is based on approximation with high estimation error, or lack of the downdate function. In this paper, a new accurate IPCA algorithm (AIPCA) which can provide both update and downdate capacity with higher accuracy because of direct accurate algebraic derivation is proposed based on the matrix additive modification. Experimental analysis is also given for evaluating the time cost and calculation accuracy of the AIPCA, the result demonstrates that the proposed method has high calculation accuracy and acceptable time consuming.