Multimodal independent component analysis - A method of feature extraction from multiple information sources

Electronics and Communications in Japan, Part III: Fundamental Electronic Science (English translation of Denshi Tsushin Gakkai Ronbunshi)

Volumn 84, Issue 11, 2001, Pages 21-28

  Akaho, Shotaro ^a   Umeyama, Shinji ^a  

^a ELECTROTECHNICAL LABORATORY   (Japan)

Author keywords

Canonical correlation analysis; Gram Charlier expansion; Independent component analysis; Multimodal data; Mutual information; Natural gradient

Indexed keywords

CORRELATION METHODS; INDEPENDENT COMPONENT ANALYSIS; INFORMATION ANALYSIS; MATHEMATICAL TRANSFORMATIONS;

CANONICAL CORRELATION ANALYSIS (CCA);

FEATURE EXTRACTION;

EID: 0034912420     PISSN: 10420967     EISSN: None     Source Type: Journal    
DOI: 10.1002/ecjc.1045     Document Type: Article

References (15)

1. Concept acquisition from multiple information sources by the EM algorithm
2. Natural gradient works efficiently in learning
3. An approximation of nonlinear canonical correlation analysis by multilayer perceptrons
4. Mutual information maximization: Models of cortical self-organization
5. The 'independent components' of natural scenes are edge filters
6. Extended ica removes artifacts from electroencephalographic recordings
7. Adaptive online learning algorithms for blind separation: Maximum entropy and minimum mutual information