Experimental Comparison between Differential Evolution and Artificial Bee Colony Algorithm: A Case Study with Continuous Optimization Problems

Abstract

This paper conducts an experimental comparison between two recently introduced meta-heuristic algorithms, which are the Differential Evolution (DE) and the Artificial Bee Colony (ABC) algorithm. Both these algorithms are very prominent and significant to represent the broader family of algorithms to which they belong, i.e., the Evolutionary and Swarm Intelligence algorithm families. Both DE and ABC have been successfully employed to numerous and diverse problems from the fields of mathematics, science and engineering. DE is an evolutionary algorithm that computes the vector differences between randomly picked candidate solution vectors and uses these differences to produce new, improved candidate solutions to advance its evolutionary search and optimization process. The ABC is a swarm intelligent algorithm that mimics the candidate solutions as a swarm of bees that forage across a search space for continuously better quality food sources (i.e., candidate solutions). The aim and focus of this paper is to present a side-by-side comparison of these two evolutionary and swarm intelligence algorithms on a common set of continuous benchmark problems to achieve a better understanding of their strengths, weaknesses and characteristics. The experimental results show that ABC is more explorative and can consistently avoid the local optima to locate the neighborhood of the global minimum, while DE is more exploitative to achieve an excellent level of fine tuning, but at the risk of premature convergence because of its lack of explorative characteristics.