Problem Decomposition in Cartesian Genetic Programming

Scalability has become a major issue and a hot topic of research for the GP community, as researchers are moving on to investigate more complex problems. Throughout nature and conventional human design principles, modular structures are extensively used to tackle complex problems by decomposing them into smaller, simpler subproblems, which can be independently solved. Modularity is defined as the degree to which an entity can be represented in terms of smaller functional blocks. These smaller functional blocks are known as modules. In this chapter, a new approach called Embedded CGP (ECGP), is described that is capable of dynamically acquiring, evolving, and reusing modules to exploit modularity. Alternative approaches for acquiring modules within ECGP are also discussed before describing Modular CGP (MCGP), an enhancement to ECGP that allows the use of nested modules to see if further performance improvements are possible. Finally, an approach that uses the concept of multiple chromosomes in order to allow CGP and ECGP to exploit modularity through compartmentalization is described.