Title :
A general framework for SVD flows and joint SVD flows
Author_Institution :
Brain Sci. Inst., RIKEN, Saitama, Japan
Abstract :
The paper presents a general framework for the development of continuous algorithms for SVD and joint SVD problems. The framework for SVD is derived based on gradient flows on unitary groups. Two previous examples of SVD flows discovered heuristically are derived systematically using the framework. The framework for SVD is extended for joint SVD problems, and a new continuous algorithm for joint SVD is derived using the framework. The work described gives a clear perspective on SVD and joint SVD flows as well as a means for the development of new continuous algorithms.
Keywords :
algorithm theory; differential equations; gradient methods; random noise; signal processing; singular value decomposition; additive noise; continuous algorithm; continuous algorithms; differential equation; gradient flows; joint SVD flows; matrix differential equations; signal processing; singular values decomposition; unitary groups; Differential equations; Eigenvalues and eigenfunctions; Gradient methods; Matrix decomposition; Source separation;
Conference_Titel :
Acoustics, Speech, and Signal Processing, 2003. Proceedings. (ICASSP '03). 2003 IEEE International Conference on
Print_ISBN :
0-7803-7663-3
DOI :
10.1109/ICASSP.2003.1202461