Genetic programming with multiple initial populations generated by simulated annealing

Abstract

Genetic Programming (GP) and Simulated Annealing Programming (SAP) are typical metaheuristic methods for automatic programming. We propose a new method, Parallel - Genetic and Annealing Programming (P-GAP) which combines GP and SAP. In P-GAP, multiple initial populations are generated by SAP. Respective populations evolve by parallel GP. As the generation proceeds, populations are integrated gradually. To examine the effectiveness, we compared P-GAP with the conventional methods in five test problems. As a result, P-GAP showed better performance than GP and SAP.