MEMETIC ALGORITHM WITH MULTI-PARENT CROSSOVER (MA-MPC) FOR MULTI-OBJECTIVE NETWORK DESIGN

Abstract

In many Evolutionary Algorithms (EAs), a crossover with two parents is commonly used to produce offsprings. Interestingly, we need not restrict ourselves to two-parent crossover since EA allows us to emulate natural evolution in a more flexible fashion. There are experimental results in the literature which show that multi-parent crossover operators can achieve better performance than traditional two-parent versions. However, most of these experimental results are based on common test functions. Experimental studies involving real-life, NP-hard problems such as network design problem are very rare. This paper presents Memetic Algorithm with Multi-Parent Crossover (MA-MPC) with a view to providing a case study of multi-parent crossover within the framework of MA for network topology design problem. Results show that MA-MPC does not always outperform MA. It depends on the size of the problem and the number parents (be it 3, 5, 7, or any other)