My doctoral adviser Mary Kurz is third author on an interesting paper in the February 2010 issue of Computers and Operations Research that compares genetic algorithms, simulated annealing, and tabu search for choosing an optimal product line. I have always been interested in the use of metaheuristics in optimization and particularly in research that compares metaheuristics or discusses how to choose and tune metaheuristics for specific problems.
My doctoral research involved machine scheduling with a random keys genetic algorithm, but I also developed a tabu search to compare with the GA. From that experience, I learned that fine-tuning a metaheuristic can really make a difference, but also that some metaheuristics may just be better suited to certain types of problems. I still lean towards using genetic algorithms because I like their multiple solution approach, but I think simulated annealing and tabu search are also very valuable tools for operations researchers.
Do you have a favorite metaheuristic? What have your experiences been with choosing and tuning metaheuristics?