Statistical Comparison of Evolutionary Algorithms
Mathieu Barrette, Bruno De Kelper, Tony Wong · ASME Press eBooks · 2007
Benchmarking evolutionary algorithm performance is an important task in computational intelligence research. However evolutionary algorithms are stochastic processes and to obtain significant results proper statistical tools must be used. This paper proposes a generic comparison procedure to assess the relative performance of several evolutionary algorithms. The proposed procedure is based on a simple ranking scheme and is statistically relevant under a control risk of error. A useful feature of the procedure is the ability to visualize the dynamical behavior of the algorithms along with the confidence intervals of their performance. It also has the advantage of eliminating the use of...