PCA Learning Algorithms with Adaptive Learning Rates

In PCA learning algorithms, learning rate is crucial. It not only influences the convergence speed of PCA algorithms but also determines whether this algorithm is convergent. Clearly, the DCT method requires that the learning rate approaches to zero. In Chapter 2, by using DDT method, it is shown that Oja’s and Xu’s LMSER’s PCA algorithms with constant learning rates are locally convergent.