Sequential Decision Making in Heuristic Search
The order in which nodes are explored in a (depth-first) iterative deepening search strategy is principally determined by the condition under which a path of the search tree is cut off in each search phase. A corresponding criterion, which has a strong influence on the performance of the overall (heuristic) search procedure, is generally realized in the form of an upper cost bound. In this paper, we develop an effective and computationally efficient termination criterion based on statistical methods of change detection. The criterion is local in the sense that it depends on properties of a path itself, rather than on the comparison with other paths. Loosely speaking, the idea is to take a systematic change in the (heuristic) evaluation of nodes along a search path as an indication of suboptimality. An expected utility criterion which also takes the consequence of the suboptimal search decision on the solution quality into account is proposed as a generalization of this idea.
KeywordsChange Detection Heuristic Search Nuisance Parameter Search Path Search Phase
Unable to display preview. Download preview PDF.
- 1.M. Basseville and I.V. Nikiforov. Detection of Abrupt Changes. Prentice Hall, 1993.Google Scholar
- 3.I. Durdanovic, H. Kleine Büning, and M. Suermann. New aspects and applications in the field of resource-based configuration. Technical Report tr-rsfb-96–023, Department of Computer Science, University of Paderborn, 1996.Google Scholar
- 4.J. Gaschnig. A problem similarity approach to devising heuristics: First results. In: Proceedings IJCAI-79, 6th International Joint Conference on Artificial Intelligence, pages 301–307, Tokyo, 1979.Google Scholar
- 5.O. Hansson and A. Mayer. Heuristic search as evidential reasoning. In: Proceedings 5th Workshop on Uncertainty in AI, pages 152–161, Windsdor, Ontario, 1989. Morgan Kaufmann.Google Scholar
- 7.M. Heinrich. Ressourcenorientiertes Konfigurieren. Künstliche Intelligenz, 193: 11–14, 1993.Google Scholar
- 8.E.J. Horvitz. Reasoning under varying and uncertain resource constraints. In: Proceeedings AAAI-88, St. Paul, MN, 1988.Google Scholar
- 9.E. Hüllermeier. Approximating cost functions in resource-based configuration. Technical Report tr-rsfb-98–060, Department of Mathematics and Computer Science, University of Paderborn, September 1998.Google Scholar
- 10.E. Hüllermeier and C. Zimmermann. A two-phase search method for solving configuration problems. Technical Report tr-rsfb-98–062, Department of Mathematics and Computer Science, University of Paderborn, October 1998.Google Scholar
- 11.E. Hüllermeier and C. Zimmermann. A two-phase search method for solving configuration problems. Technical Report tr-rsfb-98–062, Department of Mathematics and Computer Science, University of Paderborn, October 1998.Google Scholar
- 15.J. Pearl. Heuristics: Intelligent Search Strategies for Computer Problem Solving. Addison-Wesley, 1984.Google Scholar
- 16.I. Pohl. First results on the effect of error in heuristic search. In: B. Meltzer and D. Michie, (Eds), Machine Learning 5, pages 219–236. American Elsevier, New York, 1970.Google Scholar
- 17.A.E. Prieditis. Machine discovery of effective admissible heuristics. Machine Learning, 12: 117–141, 1993.Google Scholar