Exponential improvement of efficient backtracking
The paper presents a method of mechanical deduction along the lines indicated in . Attempts to find refutation (s) are recorded in the form of triples: plan, constraints, conflicts. A plan corresponds to a portion of AND/OR graph search space and represents purely deductive structure of derivation. Constraints form a graph recording the attempts of unification, while con flicts identify minimal subset of the plan, removal of which restores unifiability.
This method can be applied to any initial base of (nonnecessarily Horn) clauses. Unlike the exhaustive (blind) backtracking which treats all the goals deduced in the course of proof as equally probable source of failure, this approach detects the exact source of failure.
In this method only a small fragment of solution space is kept on disk as as a collection of triples. The search strategy and the method of non-redundant processing of individual triples which leads to a solution (if it exists) is presented. This approach is compared — on a special case — with blind backtracking and an exponential improvement is demonstrated.
Unable to display preview. Download preview PDF.
- M. Bruynooghe, L. M. Pereira, “Reivision of Top-Down Logical Reasoning Through Intelligent Backtracking”, Centre de Informatica da Universidade Nova de Lisboa, 1981.Google Scholar
- C. L. Chang, R. C. T. Lee, “Symbolic Logic and Mechanical Theorem Proving”, New York, Academic Press, 1973.Google Scholar
- P. T. Cox, T. Pietrzykowski, “Deduction Plans: A Basis for Intelligent Backtracking”, IEEE Trans. on Pattern Analysis and Machine Intelligence, Vol 3 N. 1, Jan. 1981.Google Scholar
- P. T. Cox, “Representational Economy in Mechanical Theorem-Prover”, Proc. 4th Workshop Automat. Deduction, Austin TX Feb. 1979.Google Scholar
- D. W. Loveland, “A Linear Format for Resolution”, Lect. Notes on Mathematics 125 (Symposium on Automatic Demonstration), Berlin, Spring Verlag 1970.Google Scholar
- D. Luckham, “Refinement Theorems in Resolution Theory”, Lect. Notes on Mathematics 125 (Symposium on Automatic Demonstration), Berlin, Spring, Verlag 1970.Google Scholar
- S. Matwin, T. Pietrzykowski,“Plan based Deduction: Data structure and Implementation”, this volume.Google Scholar
- L. M. Pereira, F. C. N. Pereira, H. D. Warren, “User's Guide to DEC System-10 Proglog”, Laboratoria Nacional de Engenharia Civil, Lisbon.Google Scholar
- L. M. Pereira, A. Porto, “Intelligent backtracking and Sidetracking in Horn clause programs — the theory”, Departmento de Informatica, Universidade Nova de Lisboa, 1979.Google Scholar
- L. M. Pereira, A. Porto, “Selective Backtracking for Logic Programs”, Lect. Notes in Computer Science 81 (5th Conference on Automated Deduction), Springer-Verlag 1980.Google Scholar
- T. Pietrzykowski, S. Matwin, “Linear Complexity of Efficient Backtracking in Plan-Based Deduction”, in preparation.Google Scholar
- G. Roberts, “Waterloo Prolog User's Manual”, University of Waterloo, 1980.Google Scholar
- P. Roussel, “Prolog: manuel de reference et d'utilisation”, Groupe d'Intelligence Artificielle, Marseille-Luminy, 1975.Google Scholar
- D. H. D. Warren, “Implementing Prolog”, Department of Artificial Intelligence, Edinburgh University, 1976.Google Scholar
- N. K. Zamov, V. I. Sharonov, “On a class of strategies which can be used to establish decidability of the resolution principle”, Issled po konstruktivnoye matematikye i matematicieskaye logikye III, vol 16, pp 54–64, 1969.Google Scholar