Memory and Learning in Metaheuristics

The rapid increase of dimensions and complexity of real life problems makes it more difficult to find optimal solutions by traditional optimization methods. This challenge requires intelligent and sophisticated algorithms to make the right decisions given a set of inputs and a variety of possible actions. In the problem solving arena, this definition is transformed into the term of artificial intelligence. Artificial intelligence emerges in metaheuristics via memory and learning in algorithms. Metaheuristics are promising approaches that can find near-optimal solutions in an acceptable amount of time. Many successful metaheuristics employ “intelligent” procedures to obtain high quality solutions for discrete optimization problems. To demonstrate the contribution of memory and learning into metaheuristics, Estimation of Distribution Algorithms will be incorporated as a memory and learning mechanism into Meta-RaPS (Metaheuristic for Randomized Priority Search) which is classified as a memoryless metaheuristic. The 0-1 multidimensional knapsack problem will be used to evaluate the “intelligence” of the new algorithm.