PARSE TREE POSITION MEASURING IN DISTRIBUTED GENETIC PROGRAMMING

Abstract

Distributed genetic programming (GP) is a step forward in optimization of the GP algorithm, but it suffers from the difficulties of setting the proper distribution parameters. One of the most important parameter – classes, responsible for the migration among subpopulations, can be put under the control of flocking. The challenge in applying flocking to distributed GP lies in measuring the positions and distances between the program parse trees. This paper discusses the details of possible methods for measuring the tree position, paying the most attention to resulting distance values that are of the primary goal for a successful combination of distributed GP and flocking.