Exploration of Metaheuristics through Automatic Algorithm Configuration Techniques and Algorithmic Frameworks
Alberto Franzin, Thomas Stützle · 2016
In this paper we argue that flexible algorithm frameworks can be useful to capture the wide variety of algorithmic components for heuristic algorithms and serve as basic experimental frameworks. One of the utilities is that they can implement the wide variety of different algorithm components and their alternative choices for single stochastic local search methods and we are currently extending existing frameworks in that direction. We exemplify this approach considering the example of Simulated Annealing (SA). In fact, a wide variety of design choices of SA algorithms has been proposed in the literature and algorithm frameworks may (i) simply collect potentially all available choices, (ii) provide a tool for the experimental analysis of specific algorithms and component choices, and (iii) allow the generation of new algorithm variants by combining exisiting components in new ways. We show some limited computational experiments that show the benefit of tuning in this context and the way conclusions on the performance of algorithms are altered in this way.