Global optimization, differential evolution, adaptation and self-adaptation, experimental comparison.

Seven adaptive variants of differential evolution algorithm were compared experimentally in six scalable benchmark problems at low, medium, and higher level of dimension. Four of tested algorithm variants are considered the state-of-the-art, three variants have appeared in literature recently. One of the state-of-the-art (JADE) and one other variant with adaptation based on the competition of strategies were the best performing algorithms in the test. The recommendations for application to the real-world optimization problems are derived from these experimental results.