Non-Holonomic constraints in learning blind source separation

Recently a number of adaptive learning algorithms have been proposed for blind source separation. Although the underlying principles and approaches are different, most of them have very similar forms. The algorithms are based on gradient decsent on some manifolds. In this paper, Non-Holonomic constraints are introduced and orthogonal natural gradient decsent algorithm is proposed. It is proved theoretically as well as by simulations that this algorithm has some advantage over other Holonomic constraints.