Smooth Uniform Crossover with Smooth Point Mutation in Genetic Programming: A Preliminary Study
Jonathan Page, Riccardo Poli, William B. Langdon · 1999
. In this paper we examine the behaviour of the uniform crossover and point mutation GP operators [12] on the even-n-parity problem for n = 3; 4; 6 and present a novel representation of function nodes, designed to allow the search operators to make smaller movements around the solution space. Using this representation, performance on the even-6-parity problem is improved by three orders of magnitude relative to the estimate given for standard GP in [5]. 1 Introduction Although a mutation operator is defined, the canonical form of Genetic Programming (GP) [4] relies almost exclusively on the crossover operator for exploring the solution space. GP crossover selects a random subtree from one parent program and splices it to a random location in another, affording GP the ability to search a space of arbitrary-sized programs. Its insensitivity to position in cutting and splicing (aside of satisfying constraints on tree depth), combined with the fact that multiple instances of functi...