ICA algorithms
Information Maximization approach (Tony Bell)
Blind separation of recordings in a real environment (Te-Won Lee)The ICA formulation can be extended to separated mixtures of convolved and time-delayed sources. The goal is to extract sources from a mixture which were recorded in a real environment such as an office room or conference room. Read papers and listen to some audio-demos.
Extended Infomax Algorithm (Te-Won Lee, Mark Girolami)This algorithm is able to blindly separate mixed signals with sub- and super-Gaussian source distributions. This was achieved by using a simple type of learning rule first derived by (Girolami, 1997) by choosing negentropy as a projection pursuit index. Here we use a general stability analysis to switch between sub- and super-Gaussian regimes. The algorithm can separate a variety of source distributions and is effective at separating artifacts such as eye blinks and line noise from EEG recordings. See our paper.
ICA Mixture Model (Te-Won Lee, Michael Lewicki)An extension of ICA using EM for unsupervised classification. The algorithm finds the independent sources, the mixing matrix for each class and also computes the class membership probability for each data point. This method is well suited to modeling structure in high-dimensional data and has many potential applications. See our papers.
Overcomplete ICA (Michael Lewicki)In an overcomplete basis, the number of basis vectors is greater than the dimensionality of the input. The representation of an input is not a unique combination of basis vectors, however, overcomplete representations have greater robustness in the presence of noise, are more sparse, and have greater flexibility in matching structure in the data.
