Homogeneous and Heterogeneous Island Models for the Set Cover Problem

1) Motivationg ◦Set Cover Given a set S with m elements and a set {Ci} of n subsets of S each one having a cost ci, the set cover problem requires to find a selection of subsets of S that cover the whole set and which have minimum cost. ◦GSEMO multiobjective EA on a 2-objective function (cost, uncovered elements). From Friedrich et al. (2010) approximation ratio O(logm) in polynomial time. How to make it parallel? ◦Our contribution Analysis of the runtime of two different parallel EAs, running on μ processors, to solve SetCover with the same approximation ratio as GSEMO. Comparison of different topologies and migration strategies. 2) Measuresg ◦Parallel runtime: the number of generations until at least one island has found a satisfactory solution ◦Communication effort: the total number of individuals sent between islands, throughout a run of an island model. ◦Speedup: the rate between the expected parallel runtime of the island model and the expected runtime of the same EA using only a single island.