In this paper, we proposed a novel DE variant named DE-NPC for real parameter single objective optimization. In DE-NPC algorithm, a novel adaptation scheme for the scale factor $F$ is… Click to show full abstract
In this paper, we proposed a novel DE variant named DE-NPC for real parameter single objective optimization. In DE-NPC algorithm, a novel adaptation scheme for the scale factor $F$ is first proposed, which is based on the location information of the population rather than the fitness difference. The adaptation scheme of crossover rate $CR$ in our DE-NPC is based on its success probability. Furthermore, a novel population size reduction scheme is also employed in DE-NPC, which can get a better perception of the landscape of objectives and consequently obtain an overall better performance. The algorithm validation is conducted under our test suite containing 88 benchmarks from CEC2013, CEC2014 and CEC2017 in comparison with several state-of-the-art DE variants. The experiment results show that our novel DE-NPC algorithm is competitive with these state-of-the-art DE variants.
