An Estimation of Distribution Algorithm Utilizing Opposition-Based Learning for Nonlinear Blind Sources Separation

An estimation of distribution algorithm utilizing opposition-based learning is firstly proposed in this paper. In the proposed algorithm, opposite population is generated from the current population by calculating opposite numbers, and the best individuals in the population with the current population and the opposite population are selected to form the next population based on fitness values. Then, demixing system of the blind sources separation with post-nonlinear mixture is modeled using a multi-input multi-output wavelet neural network whose parameters can be determined under the criterion of independence of its outputs. A criterion of independence based on higher order moments is used to measure the statistical dependence of the outputs of the demixing system, and the proposed algorithm is utilized to minimize the criterion. Finally, the proposed algorithm is compared with the version of the original estimation of distribution algorithm on some well-known benchmarks, and used to a post-nonlinear blind sources separation problem with two independent random signals. The relative experimental results demonstrate that the algorithm outperforms the original estimation of distribution algorithm, and is effective for post-nonlinear blind source separation.