Fitting with differential evolution: an introduction and evaluation

Differential evolution is a global optimization algorithm that has started to find widespread use in the scattering community because of its proven effectiveness. In this article the performance of the algorithm is evaluated by fitting an X-ray reflectivity data set and investigating its convergence behavior as a function of its tuning parameters. The results offer important insights for applying differential evolution algorithms to scattering problems and provide some rules of thumb on how to tune the parameters. It is shown that, by choosing optimal tuning parameter values, the speed of the fitting process can be increased by an order of magnitude.