Monotonic Convergence of a Nonnegative ICA Algorithm on Stiefel Manifold
When the independent sources are known to be nonnegative and well-grounded, which means that they have a non-zero pdf in the region of zero, a few nonnegative independent component analysis (ICA) algorithms have been proposed to separate these positive sources. In this paper, by using the property of skew-symmetry matrix, rigorous convergence proof of a nonnegative ICA algorithm on Stiefel manifold is given. And sufficient convergence conditions are presented. Simulations are employed to confirm our convergence theory. Our techniques may be useful to analyze general ICA algorithms on Stiefel manifold.
KeywordsIndependent Component Analysis Learning Rate Convergence Condition Independent Component Analysis Blind Signal
Unable to display preview. Download preview PDF.
- 6.Nishimori, Y.: Learning Algorithm for ICA for by Geodesic Flows on Orthogonal Group. In: Proc. Int. Joint. Conf. Neural Networks, Washington, DC, vol. 2, pp. 933–938 (1999)Google Scholar
- 8.Oja, E., Plumbley, M.D.: Blind Separation of Positive Sources using Nonnegative PCA. In: Proc. Int. Conf. on Independent Component Analysis and Blind Signal Separation (ICA 2003), pp. 11–16 (2003)Google Scholar