Adaptive rank-2 update algorithm for eigenvalue decomposition
G. Li, Kai‐Bor Yu · 2003
An efficient algorithm for successive eigenvalue decomposition (EVD) with adding new data to, and deleting old data from, the data matrix simultaneously is proposed. The algorithm involves converting a full-scale EVD problem by using Householder transformation twice. Some approaches to solve the resulting EVD problem, including a direct method to compute the eigenvector directly, are discussed. This algorithm can be useful in adaptive array processing and tracking of nonstationary sinusoids. >