| On 2003-11-20 at 15:00:00 (Brussels Time) |
Abstract
F-Race is a method developed by Birattari et al. to select the best configuration of a metaheuristic for a given instance class of a combinatorial optimization problem. In this presentation, I expose my use of F-Race for the selection of configurations of the iterated local search metaheuristic for a range of scheduling problems. Moreover, I propose an extension to F-Race which forces the procedure to consider the run time of the configurations in addition to the solution quality yielded on the problem instances. That is, I propose an adaptation of F-Race to make it fit for the selection of configurations for bi-objective optimization problems. Finally, I'll make an attempt to impose my views on the usefulness of survival analysis in the context of F-Race applications.
Keywords
metaheuristics, scheduling, racing, survival analysis
