Introducing the Fourth Generation of Memetic Algorithms

Memetic Algorithms (MA) are widely used heuristic algorithms designed to solve combinatorial, continuous, and multi-criteria optimization problems. The classification of memetic algorithms emphasizes three generations of MA. The first generation represents traditional hybrid algorithms, where a global optimization method is combined with one or another local search method. The second generation is usually denoted as multi-memetic algorithms and includes mechanisms of transferring memetic information and selection of memes. The third generation contains algorithms based on the self-adaptation and self-generation. The fourth generation was previously discussed but no algorithms were proposed as they would imply the generation of the algorithms during the evolution process. In this study we propose the first algorithm from the fourth generation of MA with the incorporated artificial intelligence (AI) features used for meme generation. The proposed algorithm was evaluated with a use of CEC’17 benchmark optimization functions and compared with some memetic algorithms of the previous generations.