Abstract
Admissible heuristics are an important class of heuristics worth discovering: they guarantee shortest path solutions in search algorithms such asA* and they guarantee less expensively produced, but boundedly longer solutions in search algorithms such as dynamic weighting. Unfortunately, effective (accurate and cheap to compute) admissible heuristics can take years for people to discover. Several researchers have suggested that certain transformations of a problem can be used to generate admissible heuristics. This article defines a more general class of transformations, calledabstractions, that are guaranteed to generate only admissible heuristics. It also describes and evaluates an implemented program (Absolver II) that uses a means-ends analysis search control strategy to discover abstracted problems that result in effective admissible heuristics. Absolver II discovered several well-known and a few novel admissible heuristics, including the first known effective one for Rubik's Cube, thus concretely demonstrating that effective admissible heuristics can be tractably discovered by a machine.
Article PDF
Similar content being viewed by others
References
Amarel, S. (1968). On the representation of problems of reasoning about actions. InMachine intelligence (Vol. 3). Edinburgh, Scotland: Edinburgh University Press.
Bhatnagar, N., & Mostow, J. (1990). Adaptaive search by explanation-based learning of heuristic censors. InProceedings of the Eighth National Conference on Artificial Intelligence. Boston, MA: American Association for Artificial Intelligence.
Chenoweth, S., & Davis, H. (1991). High-performance a* search with rapidly growing heuristics. InProceedings IJCAI-12. Sydney, Australia: International Joint Conferences on Artificial Intelligence.
Christensen, J., & Korf, R. (1986). A unified theory of heuristic evaluation functions and its applications to learning. InProceedings AAAI-86 (pp. 148–152). Philadelphia, PA: American Association for Artificial Intelligence.
Ellman, T. (1988). Approximate theory formation: an explanation-based approach. InAAAI88 (pp. 570–574). St. Paul, MN: American Association for Artificial Intelligence.
Ernst, G., & Goldstein, M. (1982). Mechanical discovery of classes of problem-solving strategies.JACM, 29(1), 1–23.
Fikes, R., Hart, P., & Nilsson, N.J. (1972). Learning and executing generalized robot plans.Artificial Intelligence, 3(4), 251–288. Also in B.L. Webber & N.J. Nilsson (Eds.),Readings in Artificial Intelligence. Palo Alto, CA: Tioga.
Gaschnig, J. (1979). A problem-similarity approach to devising heuristics. InProceedings IJCAI-6 (pp. 301–307). Tokyo, Japan: International Joint Conferences of Artificial Intelligence.
Guida, G., & Somalvico, M. (1979). A method for computing heuristics in problem solving.Information Sciences, 19, 251–259.
Guvenir, H., & Ernst, G.W. (1990). Learning problem solving strategies using refinement and macro generation.Artificial Intelligence, 44(1–2), 209–243.
Hansson, O., Mayer, A., & Yung, M. (in press). Criticizing solutions to relaxed models yields powerful admissible heuristics.Information Sciences.
Held, M., & Karp, R. (1970). The traveling salesman problem and minimum spanning trees.Operations Research, 18, 1138–1162.
Janakiraman, B., & Prieditis, A. (1992). Generating effective admissible project scheduling heuristics by abstraction and reconstitution. InProceedings of the AAAI 1992 Spring Symposium Series: Practical Approaches to Planning and Scheduling. Stanford, CA: American Association for Artificial Intelligence.
Keller, R. (1987).The role of explicit contextual knowledge in learning concepts to improve performance. Ph.D. thesis, Rutgers University, New Brunswick, NJ.
Kibler, D. (1985).Natural generation of heuristics by transforming the problem representation (Technical Report TR-85-20). Computer Science Department, UC-Irvine.
Kokar, M. (1986). Discovering functional formulas through changing representation. InProceedings AAAI-86. Philadelphia, PA: American Association for Artificial Intelligence.
Korf, R. (1985). Depth-first iterative-deepening: An optimal admissible tree search.Artificial Intelligence, 27(2), 97–109.
Langley, P. (1983). Learning effective search heuristics. InProceedings IJCAI-8. Karlsruhe, West Germany: International Joint Conferences on Artificial Intelligence.
Langley, P., Simon, H.A., Bradshaw, G., & Zytkow, J. (1987).Scientific discovery: Computational explorations of the creative process. Cambridge, MA: MIT Press.
Lawler, E., & Lenstra, L. (1984).The Traveling Salesman problem. New York: John Wiley and Sons.
Lenat, D.B. (1977). On automated scientific theory formation: A case study using the AM program. InMachine intelligence (Vol. 9). New York: Halsted Press.
Lenat, D.B. (1982). The nature of heuristics.Artificial Intelligence, 19, 189–249.
Lenat, D.B. (1983a). The nature of heuristics ii: Background and examples.Artificial Intelligence, 21, 31–59.
Lenat, D.B. (1983b). The role of heuristics in learning by discovery: Three case studies. InMachine learning, Los Altos, CA: Morgan Kaufmann.
Lenat, D.B., & Brown, J.S. (1984). Why AM and EURISKO appear to work.Artificial Intelligence, 23, 269–294.
Minton, S. (1988).Learning effective search-control knowledge: An explanation-based approach. Ph.D. thesis, Carnegie-Mellon University, Pittsburgh, PA.
Mitchell, T.M., Utgoff, P.E., & Banerji, R.B. (1983). Learning by experimentation: Acquiring and refining problem-solving heuristics. InMachine learning: An artificial intelligence approach. Palo Alto, CA: Tioga.
Moshi, A. (1989).Transformations for backtracking SETL programs. Ph.D. thesis, New York University.
Mostow, J., & Bhatnagar, N. (1987). Failsafe—a floor planner that uses EBG to learn from its failures. InProceedings of the Tenth International Joint Conference on Artificial Intelligence (IJCA187), Vol. 1 (pp. 249–255). Milan, Italy: Morgan Kaufmann.
Mostow, J. & Prieditis, A. (1989). Discovering admissible heuristics by abstracting and optimizing. InProceedings IJCAI-11, Detroit, MI: International Joint Conferences on Artificial Intelligence.
Nilsson, N.J. (1980).Principles of artificial intelligence. Palo Alto, CA: Morgan Kaufmann.
Oyen, R.A. (1975).Mechanical discovery of invariances for problem solving (Technical Report 1168). Cleveland, OH: Computer Engineering Department, Case Western University.
Paige, R., & Koenig, S. (1982). Finite differencing of computable expressions.ACM TOPLAS, 4(3), 402–454.
Pearl, J. (1984).Heuristics: Intelligent search strategies for computer problem-solving. Reading, MA: Addison-Wesley.
Pohl, I. (1973). The avoidance of (relative) catastrophe, heuristic competence, genuine dynamic weighting and computational issues in heuristic problem solving. InProceedings IJCAI-3 (pp. 20–23). Stanford, CA: International Joint Conferences on Artificial Intelligence.
Prieditis, A. (1990).Discovering effective admissible heuristics by abstraction and speedup: A transformational approach. Ph.D. thesis, Rutgers University, New Brunswick, NJ.
Rendell, L. (1976).A method for automatic generation of heuristics for state-space problems (Technical Report CS-76-10). Waterloo, Ontario: University of Waterloo, CS Department.
Samuel, A.L. (1963). Some studies of machine learning using the game of checkers. InComputers and Thought. New York: McGraw-Hill.
Schaffer, C. (1990). A proven domain-independent scientific function-finding algorithm. InProceedings AAAI-90. Boston, MA: American Association for Artificial Intelligence.
Valtorta, M. (1984). A result on the computational complexity of heuristic estimates for the A* algorithm.Information Sciences, 34, 47–59.
Żytkow, J., Zhu, J., & Hussam, A. (1990). Automated discovery in a chemistry laboratory. InProceedings AAAI-90. Boston, MA: American Association for Artificial Intelligence.
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Prieditis, A.E. Machine discovery of effective admissible heuristics. Mach Learn 12, 117–141 (1993). https://doi.org/10.1007/BF00993063
Received:
Issue Date:
DOI: https://doi.org/10.1007/BF00993063