An Algebraic Algorithm for Independent Component Analysis With More Sources Than Sensors
L De Lathauwer, B De Moor, J Vandewalle · 2002
Abstract The goal of independent component analysis now consists of the estimation of the mixing matrix and/or the corresponding realizations of the source vector X, given only realizations of the observation vector Y. The key assumption is that the components of X are mutually statistically independent, as well as statistically independent from the noise components. This is a very strong hypothesis, but also quite natural in lots of applications. It means that the aim can often be rephrased as splitting the dataset into components ‘of a different nature’, which contributed to the data in a linear way.