Title: A method for sparse and robust independent component analysis Authors:  Lauri Heinonen - University of Turku (Finland) [presenting]
Joni Virta - University of Turku (Finland)
Abstract: Independent component analysis (ICA) is a popular family of methods for decomposing signals into independent sources. One group of ICA methods are those achieved by using symmetrized scatter matrices or other scatter matrices with independent properties in invariant coordinate selection (ICS). A sparse version of this method is presented to achieve sparse independent component analysis (SICA). When using scatter matrices that are also robust, the SICA method is also robust. Compared to regular ICA, sparse ICA gives sparse loadings where some of the loadings are estimated to be exactly zero. The SICA method is presented as a sequence of regression problems, and the LASSO penalty is used to achieve sparsity. The method is illustrated with examples and compared to different ICA methods and other relevant competitors.