Blind signal separation and Independent component analysis

Course Description

Mathematical preliminaries: stochastic processes, gradients and optimization methods, information theory. Blind separation of signals by principal (PCA) and independent (ICA) component analysis: conditions for uniqueness of the solution. Information-theoretic approach to ICA. Equivalence between ICA methods based on minimum of mutual information and maximum negentropy. Single and multichannel blind deconvolution in time and frequency domains. Underdetermined blind source separation: sparse component analysis and nonnegative matrix factorizations.

Study Programmes

Postgraduate doctoral study programme

Literature

Aapo Hyvärinen, Juha Karhunen, Erkki Oja (2004.), Independent Component Analysis, John Wiley & Sons
Andrzej Cichocki, Shun-ichi Amari (2002.), Adaptive Blind Signal and Image Processing, John Wiley & Sons
T. - M. Huang, V. Kecman, I. Kopriva (2006.), Kernel based Algorithms for Mining Huge Dana Sets: Supervised, Semi-supervised and Unsupervised Learning, Springer Series: Studies in Computational Intelligence, Vol. 17, XVI

General

ID 154771
  Summer semester
6 ECTS
L3 English Level
L1 e-Learning