A Quasi-Newton Algorithm for Minor Component Analysis and Its Convergence

P.C. Ching, S. Ouyang, and T. Lee (PRC)


Subspace Tracking, Minor Component Analysis (MCA),Newton Method, Adaptation, Convergence Analysis.


A quasi-Newton algorithm for adaptive Minor Component Analysis (MCA) is introduced. The adaptation proper ties are rigorously analyzed using stochastic approximation theory, and its exponential convergence rate is established. The convergence performance and tracking capability are evaluated by computer simulations for both stationary and nonstationary scenarios. It is shown that the novel quasi Newton algorithm is effective in providing accurate MCA subspace estimate.

Important Links:

Go Back