ARP: asexual reproduction programming

Abstract

Recently meta-heuristic techniques have attracted more attention. Algorithms based on Bio-inspired problems are among the most popular techniques of this field. In meta-heuristic algorithms, Genetic algorithm is one of the most useful. GA uses chromosome representation and operates on the chromosome with crossover and mutation operators. Genetic programming is a form of GA with tree representation for its chromosomes. GP was developed to evolve programming in computers and is a population-based algorithm. But GP is very slow and needs a long time for converging. On the other hand, asexual reproduction optimisation (ARO) is another variant of meta-heuristic algorithms in which convergence to the global optima is done at a fast time. In this paper, we introduced a new method, which is inducted by asexual reproduction with combination to GP. This algorithm is named Asexual Reproduction Programming (ARP). ARP has advantages of both ARO and GP together i.e. the fast convergence time of ARO and the power and flexibility of GP. ARP has fast convergence to global optimum while its error is less than GP. By mathematically analysing and proving, we show the ARP convergence to the global optimum. To assay the efficiency of the ARP, two algorithms were compared on some real-valued symbolic regression problems. Perusing the experimental results demonstrate that ARP outperforms GP in performance and convergence time.