Topographic independent component analysis

Citation
A. Hyvarinen et al., Topographic independent component analysis, NEURAL COMP, 13(7), 2001, pp. 1527-1558
Citations number
44
Categorie Soggetti
Neurosciences & Behavoir","AI Robotics and Automatic Control
Journal title
NEURAL COMPUTATION
ISSN journal
08997667 → ACNP
Volume
13
Issue
7
Year of publication
2001
Pages
1527 - 1558
Database
ISI
SICI code
0899-7667(200107)13:7<1527:TICA>2.0.ZU;2-Z
Abstract
In ordinary independent component analysis, the components are assumed to b e completely independent, and they do not necessarily have any meaningful o rder relationships. In practice, however, the estimated "independent" compo nents are often not at all independent. We propose that this residual depen dence structure could be used to define a topographic order for the compone nts. In particular, a distance between two components could be defined usin g their higher-order correlations, and this distance could be used to creat e a topographic representation. Thus, we obtain a linear decomposition into approximately independent components, where the dependence of two componen ts is approximated by the proximity of the components in the topographic re presentation.