Regarding the rankings of optimization heuristics based on artificially-constructed benchmark functions

摘要: Novel Evolutionary Algorithms are usually tested on sets of artificially-constructed benchmark problems. Such problems often created to make the search one global extremum (usually minimum) tricky. In this paper it is shown that benchmarking heuristics either minimization or maximization same set artificially-created functions (with equal bounds and number allowed function calls) may lead very different ranking algorithms. As other heuristic optimizers developed in order be applicable real-world problems, such result raise doubts practical meaning them artificial functions, as there little reason searching for minimum should more important than their maximum.Thirty optimization heuristics, including a variants Differential Evolution, well kinds Algorithms, Particle Swarm Optimization, Direct Search methods - following idea borrowed from No Free Lunch pure random paper. Some discussion regarding choice mean median performance comparison addressed short debate overall particular given.