Comparison of Evolutionary Multi-Objective Optimization Algorithms for the utilization of fairness in network control

Mario Köppen, Rodrigo Verschae, Kaori Yoshida, Masato Tsuru

Abstract: We use design principles of evolutionary multi-objective optimization algorithms to define algorithms capable of approximating maximum sets of relations in general. The specific case of fairness relations is considered here, which play a prominent role in the control of resource sharing in data networks. We study maxmin fairness allocation in networks with linear congestion control. Among various design principles, the concepts behind Strength Pareto Evolutionary Algorithm, and the Multi-Objective Particle Swarm Optimization achieve comparable best performance (with the used parameterization within 10% of the fairness state components for up to 20 objectives).