Nonlinear blind separation algorithm using multiobjective evolutionary algorithm

In nonlinear blind source separation, the approach for invertible functions is very difficult due to the existence of many local minima. For separating source signals efficiently, a specific-designed multi-objective evolutionary algorithm is proposed. As defining a novel kind of multiple fitness functions by the maximum value of the normalized objective multiplied by weights, the evolutionary algorithm can explore the search space uniformly, keep the diversity of the population, and escape from local optima. The simulation results demonstrate that the proposed algorithm is efficient.