Constrained ICA

We present the code for the non-orthogonal constrained extended Infomax (CD-Infomax) algorithm as described in [1]. This algorithm allows constraints on either the latent sources or the mixing matrix to be added to the ICA cost function and the optimization is performed in the Lagrangian framework. In addition, the extended Infomax algorithm is able to estimate a more diverse set of sources by using two different nonlinearities to estimate both super- and sub-Gaussian distributions [2].


    [1] P. A. Rodriguez, M. Anderson, X-L Li, and T. Adali, "General Non-Orthogonal Constrained ICA," IEEE Transactions on Signal Processing, vol.62, no.11, pp.2778-2786, June 2014.
    [2] Lee, Te-Won, Mark Girolami, and Terrence J. Sejnowski, "Independent Component Analysis Using an Extended Infomax Algorithm for Mixed Sub-Gaussian and Super-Gaussian Sources," Neural Computation, vol. 11, no. 2, pp.417-441, Feb. 1999.