An Individual-based Approach To Multi-level Selection

Evolutionary Algorithms (EAs) are in this case counter-intuitive since they try to evolve a solution for the entire problem as a whole. EAs may show improvement when they can create more complex evolutionary units through some form of cooperative combination of sub-solutions similar to divide-and-conquer. In other words, instead of trying to evolve a single solution for the problem, solutions may try to cooperate to solve the problem. This cooperation results in groups of individuals which have more functionality than an isolated individuals. Hence, complexity, in terms of the structure of the solution, becomes an emergent property of the evolutionary system.