ABSTRACT
Principal component analysis (PCA) neural networks are useful tools in feature extraction, data compression, pattern recognition, and time series prediction, especially in online data precessing applications. The learning algorithms of PCA neural networks play an important role in their practical applications. Stemming from Oja’s algorithm [134], many PCA algorithms have been proposed to update the weights of these networks. Among the algorithms for PCA, Oja’s algorithm and Xu’s LMSER algorithm are commonly used in practical applications. Several other algorithms for PCA are related to the two basic procedures [32]. In this chapter, we will study the Oja’s and Xu’s algorithms and establish the theoretical foundation of their applications.
