23rd annual Belgian-Dutch Conference on Machine Learning pages:48-54
Benelearn edition:23 location:Brussels date:06 June 2014
Many methods for multi-objective optimisation exist, and there are multiple studies in which their performance is compared in terms of a wide range of evaluation metrics. Usually, these studies compare the end result of the optimisation process on given benchmarks; they do not evaluate how fast this end result is obtained, nor how properties of the benchmarks affect these results. In this paper, we investigate how the search space dimensionality of optimisation problems affects the behaviour of different methods, not only in terms of the end result but also in terms of how fast it is achieved. We compared two particle-swarm based optimisers, an elitist evolutionary algorithm and a scatter search algorithm. Our results show that while the PSO-based methods generally converge faster or equally fast compared to the others, they found a less diverse set of solutions.