JADE with k Nearest Neighbors Surrogate Model

JADE belongs to the top efficient versions of the Differential Evolution concept. We couple JADE with k Nearest Neighbors surrogate model approximating the fitness function. The model is global, i.e., it is identical for all points in the population. The procedure to control the surrogate model learning process is a version of a method that has been initially introduced for the CMA-ES algorithm. Tests of the algorithm on the CEC'2013 benchmark suite reveal the efficiency increase thanks to the surrogate model for a majority of problems contained in the test suite.