Discrete optimization problemsarise throughout many real world domainsincluding planning, decision making, and search. NP-hardin general, these problems require novel approachesto algorithmdesign and development.Algorithmportfolios whichconsist of multiple algorithminstances and types executingin parallel has been a key novel approachto these problems. \Ve focus oil cooperative portfolios wherealgorithms actively communicateand interact with one another during problemsolving to moreefficiently determine the optimal solution. Unfortunately, the addeddimension of cooperation greatly complicates our understanding of tile method. Thegoal of this paper is to provide tile first comprehensiveformal computationalmodelof portfolios of cooperative heterogeneousalgorithms in order to lay the foundations for rigorous analyses of these portfolios in hopesof eventuallyleading to provable guaranteesof performance,effectiveness, and efficiency.
Eugene Santos Jr.