Two approaches to estimation of overcomplete independent component bases
Mika Inki, Aapo Hyvärinen · 2003
Estimating overcomplete ICA bases is a difficult problem that emerges when using ICA on many kinds of natural data. Here we introduce two algorithms that estimate an approximate overcomplete basis quite fast in a high-dimensional space. The first algorithm is based on an assumption that the basis vectors are randomly distributed in the space, and the second on the gaussianization procedure.