Niching the CMA-ES via nearest-better clustering

We investigate how a niching based evolutionary algorithm fares on the BBOB function test set, knowing that most problems are not very well suited to this algorithm class. However, as the CMA-ES is included as basic local search algorithm, the niching approach still performs fairly well, with some potential to improve. Basin identification is done via the heuristic nearest-best clustering scheme. Categories and Subject Descriptors

• Type: Conference
• Year: 2010
• Where: GECCO
• Authors: Mike Preuss
• Keywords: Algorithm Class, BBOB Function Test, Evolutionary Algorithm Fares, GECCO 2010, Optimization