Cooperative Coevolution Framework with Recursive Differential Grouping Algorithms for Solving Large-Scale Optimization Problems
Klemen Berkovič, Borko Boškovič, Janez Brest · 2025
Cooperative coevolution is a subgenre of computational intelligence used to optimize high-dimensional problems through a divide-and-conquer approach. The main challenges of cooperative coevolution lie in problem decomposition and the use of a suitable optimization algorithm. Decomposition addresses mainly the division of the problem into smaller sub-problems, where the main challenge is determining the interactions between the components of the problem. In our work we implemented five different decomposition algorithms and developed a cooperative co-evolutionary framework. The framework uses implemented decomposition algorithms for problem decomposition and a particle swarm algorithm for optimizing the subproblems. The five decomposition algorithms are based on the recursive differential grouping algorithm. We conducted two experiments. First, we analyzed the accuracy of the implemented decomposition algorithms, and, in the second experiment, we analyzed the influence of the proposed cooperative coevolutionary framework to the particle swarm optimization algorithm. In the second experiment we used five different cooperative coevolution algorithms that used five implemented recursive differential grouping algorithms. We found that the cooperative coevolution framework can significantly improve the results of optimization compared to the particle swarm algorithm. We also found that the choice of the problem decomposition strategy plays an important role.