Parallel Genetic Programming Induction of Binary Decision Diagrams

Genetic programming is a new technique for machine learning, program induction and optimization loosely based on an evolutionary paradigm. Genetic programming is easily amenable to parallel computing which help relieve the intrinsic slowness of the approach. We describe a parallel implementation of genetic programming on the T3D computer. We apply the system to a problem of induction of binary decision diagrams used in logical circuit design. It is shown that the results depend in a critical way on the representation of the decision diagrams and that the parallel implementation is able to find the correct solution with less computational effort than the sequential version.