Independent Component Analysis Based on Information Bottleneck. (7th June 2015)

• Record Type: Journal Article
• Authors: Ke, Qiao; Zhang, Jiangshe; Srivastava, H. M.; Wei, Wei; Chen, Guang-Sheng
• Other Names: Zhang Hui Academic Editor.
• Is Part Of: Abstract and applied analysis. Volume 2015(2015)
• Publication Date: 2015-06-07
• Subjects: Mathematical analysis -- Periodicals; Mathematical analysis; Applied Mathematics; Mathematical Analysis; Periodicals; 515.05
• DOI: 10.1155/2015/386201
• Languages: English
• ISSNs: 1085-3375

Abstract

Abstract : The paper is mainly used to provide the equivalence of two algorithms of independent component analysis (ICA) based on the information bottleneck (IB). In the viewpoint of information theory, we attempt to explain the two classical algorithms of ICA by information bottleneck. Furthermore, via the numerical experiments with the synthetic data, sonic data, and image, ICA is proved to be an edificatory way to solve BSS successfully relying on the information theory. Finally, two realistic numerical experiments are conducted via FastICA in order to illustrate the efficiency and practicality of the algorithm as well as the drawbacks in the process of the recovery images the mixing images.