Binarized Monte Carlo Search for Selection Problems
摘要
In this paper, we consider adaptations of Monte Carlo Search methods on binary decision trees where actions are simulated using heuristics and where choices are made deterministically or stochastically. We explain how these adaptations are fitted for combinatorial problems such as element selection problems in order to compete with other approximate resolution methods such as metaheuristics. We present results on a theoretical problem (Set Covering) and on an applied problem (Pulse Repetition Frequency Selection) with different simulation heuristics. We then discuss the usefulness of these new methods based on the characteristics of the problems and on the quality of the simulation heuristics used to construct the decision tree.