Case-based reasoning (CBR) aims at using experience from the past in order to guide future problem solving rather than "starting from scratch" every time. We propose a CBR strategy particularly suitable for realizing this principle if heuristic search is used as a problem solving method: Given a new problem, a CBR method exploits previously solved problems in order to predict a region of the search space which is (provably) probable to contain the solution. The efficiency of a search method applied afterwards for actually finding the solution is then improved by focusing on this region. Our results provide a formal basis for the intuitively meaningful (even though not always justified) idea to concentrate on those parts of the search space where solutions to similar problems have already been found. The approach outlined in this paper either can be seen as one of CBR-supported heuristic search or as a formal framework of (search-oriented) CBR.
Keywords: Case-Based Reasoning, Search
Citation: Eyke Hüllermeier: Focusing Search by Using Problem Solving Experience . In W.Horn (ed.): ECAI2000, Proceedings of the 14th European Conference on Artificial Intelligence, IOS Press, Amsterdam, 2000, pp.50-54.