Maximum Likelihood Independent Component Analysis using GA and PSO

In this paper, the problem of independent component analysis based on maximum likelihood criteria has been considered. This approach involves inversion of matrix in each iteration (which is computationally complex). Therefore, genetic algorithm and particle swarm optimization have been proposed to be used for solving maximum likelihood ICA problem. Results, given in MSE of estimating signals with respect to source signals, show good performance of the proposed algorithms. Also a comparison to traditional FastICA method has been presented, which shows acceptable performance of the algorithms