Quality Criteria of Genetic Algorithms for Structure Optimization

A set of quality criteria is proposed to evaluate the performance of genetic algorithms for optimization. Instead of concentrating solely on the best solutions proposed by the algorithm, the quality criteria also consider the repeatability of the optimization and the coverage of the search space. They are tested by using various parameter settings of a genetic algorithm providing starting structures for molecular mechanic calculations of organic molecules. These or similar criteria can also be used for other domains of optimization with evolutionary algorithms.