View article

This document is based on the BBOB 2009 function document [1]. In the following, 24 noisefree real-parameter single-objective benchmark functions are presented. Our intention behind the selection of benchmark functions was to evaluate the performance of algorithms with regard to typical difficulties which we believe occur in continuous domain search. We hope that the function collection reflects, at least to a certain extend and with a few exceptions, a more difficult portion of the problem distribution that will be seen in practice (easy functions are evidently of lesser interest).

We prefer benchmark functions that are comprehensible such that algorithm behaviours can be understood in the topological context. In this way, a desired search behaviour can be pictured and deficiencies of algorithms can be profoundly analysed. Last but not least, this can eventually lead to a systematic improvement of algorithms.