Title :
Convergence analysis of distributed PAST based on consensus propagation
Author :
Reyes, Carlos ; Dallinger, Robert ; Rupp, Markus
Author_Institution :
Inst. of Telecommun., Vienna Univ. of Technol., Vienna, Austria
Abstract :
In previous work, we have analyzed the convergence properties of the so-called Projection Approximation Subspace Tracking algorithm (PAST) by means of Singular Value Decomposition methods. Based on this, we extend our analysis and study the global convergence properties of two distributed variants of the PAST algorithm based on average consensus. We provide an insight regarding convergence in the mean, and establish step-size bounds that guarantee the stability of the algorithm. Finally, we investigate the algorithms behavior at several step-sizes by means of simulation experiments.
Keywords :
approximation theory; singular value decomposition; PAST algorithm; consensus propagation; convergence analysis; distributed PAST; global convergence properties; projection approximation subspace tracking algorithm; singular value decomposition methods; step-size bounds; PAST algorithm; average consensus; blind source separation; distributed subspace tracking; step-size bounds;
Conference_Titel :
Signals, Systems and Computers (ASILOMAR), 2012 Conference Record of the Forty Sixth Asilomar Conference on
Conference_Location :
Pacific Grove, CA
Print_ISBN :
978-1-4673-5050-1
DOI :
10.1109/ACSSC.2012.6489005
