Fast adaptive algorithm to extract multiple principal generalized eigenvectors
Jian Yang, Xi Chen · 2011
We consider adaptively extracting multiple principal generalized eigenvectors, which can be widely applied in modern signal processing. By using deflation technique, the problem is reformulated into an unconstrained minimization problem. An adaptive sequential algorithm based on Newton method is proposed to solve this problem. In order to improve its real-time performance, a parallel version of this algorithm is provided on the basis of certain approximation. Furthermore, a two-layer neural network is constructed to execute the adaptive algorithm. The simulation results demonstrate the effectiveness of the proposed algorithms.