ABSTRACT
In this chapter, evolutionary computation is presented as a methodology for solving many current problems encountered in the neural network design process. Several design areas are addressed including alternative training methods, which prevent entrapment in local minima points, automatic selection of optimal neural topologies, and determination of optimal input feature sets. Differences between conventional (i.e. gradient-based learning algorithms, mean-squared-error optimization) and evolutionary computation approaches are discussed along with current application areas and future research directions.
