ABSTRACT
In this chapter a computing method is introduced that is based on a biological metaphor of evolution: genetic algorithms. We will briefly review its principles and discuss some of its characteristics. Among others, it will be argued that modularity plays an important role with genetic algorithms as well, although in a somewhat disguised form. We will also discuss some aspects of the relation between learning and evolution. Then, some simulation studies that combine genetic algorithms with neural networks are briefly reviewed. Finally, we will apply genetic algorithms to the design of modular neural networks. In particular, we will investigate to what extent they can be used to improve the results of the model for handwritten numerals of the previous chapter.
