Parallel hybrid evolutionary computation: Automatic tuning of parameters for parallel gene expression programming

In this talk I present a parallel hybrid framework that combines gene expression programming (GEP) as the evolutionary problem­solving methodology and alternative meta­heuristics for tuning parameter values of parallel GEP runs. The implementation of our framework is based on a client­server architecture, including clients that use GEP to evolve candidate solutions for the problem in question, and clients that use (possibly) other meta­heuristics to tune GEP input parameters. We have used a suite of symbolic regression problems of different complexities to test this framework. Our results show that our approach provides a solution for the problem of automatically tuning two GEP input parameters, viz. the number of genes and the length of each gene.