Natural gradient learning for second-order nonstationary source separation

Author

Choi, Seungjin ; Cichocki, Andrzej ; Amari, Shunichi

Author_Institution

Dept. of Comput. Sci. & Eng., POSTECH, South Korea

Volume

1

fYear

2002

fDate

6/24/1905 12:00:00 AM

Firstpage

654

Lastpage

658

Abstract

In this paper we consider a problem of source separation when sources are second-order nonstationary stochastic processes. We employ the natural gradient method and develop learning algorithms for both linear feedback and feedforward neural networks. Thus our algorithms possess equivariant property. The local stability analysis shows that separating solutions are always locally stable stationary points of the proposed algorithms, regardless of probability distributions of sources

Keywords

feedforward neural nets; gradient methods; learning (artificial intelligence); probability; signal detection; stability; stochastic processes; feedforward neural networks; learning algorithms; linear feedback neural networks; local stability analysis; locally stable stationary points; natural gradient method; nonstationary source separation; probability distributions; second-order stochastic processes; Decorrelation; Feedforward neural networks; Gradient methods; Neural networks; Neurofeedback; Output feedback; Probability distribution; Source separation; Stochastic processes; Vectors;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks, 2002. IJCNN '02. Proceedings of the 2002 International Joint Conference on

Conference_Location

Honolulu, HI

ISSN

1098-7576

Print_ISBN

0-7803-7278-6

Type

conf

DOI

10.1109/IJCNN.2002.1005550

Filename

1005550