Seminario 23/13 – Christian Blum (Artificial Intelligence Research Institute, IIIA-CSIC): Recent Developments in Hybrid Metaheuristics for Combinatorial Optimization

Información

A lot of research has been dedicated to solving combinatorial optimization problems during past decades, both in the Operations Research field and in Artificial Intelligence. As a result, researchers and/or practitioners are able to choose between a wide range of both exact and approximate techniques when faced with an optimization problem of that type. Nevertheless, producing good-enough solutions to a given problem within an acceptable computation time still requires a considerable amount of experience. Therefore, tools such as MILP solvers or metaheuristic libraries enjoy en increasing popularity. In an attempt to develop a general recipe that is rather easy to apply to a given combinatorial optimization problem, my group has developed an algorithmic approach that combines heuristic elements with the power of available MILP solvers. This algorithmic approach is known as «Construct, Merge, Solve & Adapt (CMSA)». In this talk, I will give a gentle introduction to CMSA and I will present its advantages as well as recent variants. In addition, I will present STNWeb, a new tool for the graphical comparison of the behavior of multiple algorithms applied to the same problem instance.