A Framework for Eigen and Singular Component Analysis
Mohammed A. Hasan · Proceedings of the ... American Control Conference/Proceedings of the American Control Conference · 2007
A framework that involves an unconstrained optimization of a polynomial type cost function weighted with a diagonal matrix is utilized to develop learning rules for principal and minor component analyzers. With some modifications, this cost function is also used to derive generalized principal and minor component analyzers, and principal singular component analyzers. Global and asymptotic stability of the proposed systems are analyzed via Liapunov theory and the Lasalle invariance principle.