Parallel Distributed Genetic Programming

This paper describes Parallel Distributed Genetic Programming (PDGP), a new form of Genetic Programming (GP) which is suitable for the development of programs with a high degree of paral-lelism and an eecient and eeective reuse of partial results. Programs are represented in PDGP as graphs with nodes representing functions and terminals, and links representing the ow of control and results. In the simplest form of PDGP links are directed and unlabelled, in which case PDGP can be considered a generalisation of standard GP. However, more complex (direct) representations can be used, which allow the exploration of a large space of possible programs including standard tree-like programs, logic networks, neural networks, recurrent transition networks, nite state au-tomata, etc. In PDGP, programs are manipulated by special crossover and mutation operators which guarantee the syntactic correctness of the oospring. For this reason PDGP search is very eecient. PDGP programs can be executed in diierent ways, depending on whether nodes with side eeects are used or not. The paper describes the representations, the operators and the interpreters used in PDGP, and illustrates its behaviour on a large number of problems.