Title :
A weighted logarithmic merit function for canonical correlation analysis
Author :
Hasan, Mohammed A.
Author_Institution :
Dept. of Electr.&Comput. Eng., Univ. of Minnesota Duluth, Duluth, MN
Abstract :
A weighted logarithmic merit function that incorporates a diagonal matrix is utilized for deriving a gradient dynamical system that converges to the actual canonical correlation coordinates of arbitrary data matrices. The equilibrium points of the resulting gradient system are determined and their stability is thoroughly analyzed. Qualitative properties of the proposed systems are analyzed in detail including the limit of solutions as time approaches infinity. The performance of this system is also examined.
Keywords :
gradient methods; matrix algebra; statistical analysis; arbitrary data matrices; canonical correlation analysis; diagonal matrix; equilibrium points; gradient dynamical system; matrix calculus; unconstrained optimization; weighted logarithmic merit function; Calculus; Computed tomography; Data engineering; H infinity control; Image analysis; Information analysis; Matrix decomposition; Signal analysis; Stability analysis; Symmetric matrices; Canonical correlation analysis(CCA); SVD; gradient dynamical system; matrix calculus; unconstrained optimization;
Conference_Titel :
Acoustics, Speech and Signal Processing, 2009. ICASSP 2009. IEEE International Conference on
Conference_Location :
Taipei
Print_ISBN :
978-1-4244-2353-8
Electronic_ISBN :
1520-6149
DOI :
10.1109/ICASSP.2009.4959940
