ABSTRACT
Genetic Programming is the development of computer programs by evolutionary means. Genetic Programming is able to solve an impressive variety of problems from different problem domains. This chapter presents a new modular approach to Genetic Programming which is based on the introduction of local modules. Modules are allowed to evolve at a much slower rate than programs reflecting the need of programs to rely on their modules for improving their function. One of the important issues in Genetic Programming is whether GP is able to scale up. The idea of local and context-sensitive modules is motivated by the success of F. Gruau's work on cellular encoding. At the surface, cellular encoding is about making graphs available for use with genetic programming. Gruau develops neural networks, other researchers develop other graph-like applications, for example, electric circuits. The aspect interesting here, however, is that of hierarchical evolution.
