Iteration Strategy and its Effect Towards the Performance of Population Based Metaheuristics

Metaheuristics algorithms solve optimization problems by repeating a set of procedures. The algorithms can be categorized based on number of agents, either single agent algorithms which are also known as single solution metaheuristics or multi agents algorithms, also known as population-based metaheuristics. In single solution based algorithms, the steps are executed one by one by the single search agent. However, the sequence of the procedures execution with respect to members of a population becomes an issue in population-based algorithms. This issue is governed by iteration strategy, which affects the flow of information within the population. The effect of iteration strategy is studied here. This is an important issue to be considered when designing a new population- based metaheuristic. Three parent algorithms, namely, particle swarm optimization (PSO), gravitational search algorithm (GSA), and simulated Kalman filter (SKF) are used in this work to find a general pattern of the effect of iteration strategy towards the performance of population-based algorithms. Here, the effect of iteration strategy is studied using the CEC2014's benchmark functions. The finding shows that iteration strategy can influence the performance of an algorithm and the best iteration strategy is unique to its parent algorithm. A researcher developing a new population-based algorithm need to identify the best strategy so that the performance of the algorithm proposed is maximized.