Evolutionary non-linear system identification

In this paper we are showing results in nonlinear system identification (SI) using a breeder genetic algorithm (BGA) with fuzzy recombination as optimization method. We applied this evolutionary algorithm (BGA) to static and a dynamic SI problem. In the nonlinear static case, we have that traditional methods usually needs a several step procedure for optimizing the parameters values; usually the user attempts to obtain a linear form of the model using transformations in order to perform a first parametric approximation with the Least Mean Squared method and then change to a gradient based method for accuracy. Here, we are using evolutionary algorithms for performing in one step procedure the whole optimization task. For the dynamic case, we are showing preliminary results optimizing parameters of a nonlinear finite impulse response filter (NFIR).