Abstract
Problems in automated deduction essentially amount to hard search problems. Powerful search-guiding heuristics are indispensable if difficult problems are to be handled. A promising and natural way to improve the performance of heuristics is to learn from previous experience. We present heuristics that follow this approach. A first heuristic attempts to re-enact a proof of a proof problem found in the past in a flexible way in order to find a proof of a similar problem. A further heuristic employs “features” in connection with past proof experience to prune the search space. Both heuristics not only allow for substantial speed-ups, but also make it possible to prove problems that were out of reach when using socalled basic heuristics. A combination of these two heuristics can further increase performance. We demonstrate the power of our heuristics in the light of problems stemming from the study of logic calculi in connection with an inference rule called “condensed detachment”. These problems are widely acknowledged as prominent test sets for automated deduction systems and their search-guiding heuristics. We compare our results with the results the creators of Otter obtained with this renowned theorem prover and this way substantiate our achievements.
This work was supported by the Deutsche Forschungsgemeinschaft (DFG).
Preview
Unable to display preview. Download preview PDF.
References
Avenhaus, J.; Denzinger, J.: Distributing equational theorem proving, Proc. 5th RTA, Montreal, CAN, 1993, LNCS 690, pp. 62–76
Brock, B.; Cooper, S.; Pierce, W.: Analogical reasoning and proof discovery, Proc. CADE 9, Argonne, IL, USA, 1988, LNCS 310, pp. 454–468
Denzinger, J.:Knowledge-Based. Distributed Search Using Teamwork, Proc. 1stth ICMAS, San Francisco, CA, USA, 1995, pp. 81–88
Fuchs, M.:Learning proof heuristics by adapting parameters, Proc. 12th ICML, Tahoe City, CA, USA, 1995, pp. 235–243
Fuchs, M.:Experiments in the Heuristic Use of Past Proof Experience, SEKI-Report SR-95-10, University of Kaiserslautern, 1995, obtainable via WWW at the URL http: //www.uni-kl.de/AG-AvenhausMadlener/fuchs.html
Fuchs, M.:Powerful Search Heuristics Based on Weighted Symbols, Level and Features, Proc. FLAIRS '96, Key West, FL, USA, 1996
Koehler, J.; Nebel, B.:Plan modification versus plan generation, Proc. IJ-CAI '93, Chambery, FRA, 1993, pp. 1436–1444
Kolbe, T.; Walther, C.: Reusing proofs, Proc. 11th ECAI '94, Amsterdam, HOL, 1994, pp. 80–84
Kolbe, T.; Walther, C.: Patching Proofs for Reuse, Proc. 8th ECML '95, Heraklion, Crete/Greece, 1995
Łukasiewicz, J.:Selected Works, L. Borkowski (ed.), North-Holland, 1970
McCune, W.; Wos, L.: Experiments in Automated Deduction with Condensed Detachment, Proc. CADE 11, Saratoga Springs, NY, USA, 1992, LNAI 607, pp. 209–223
Suttner, C.; Ertel, W.: Automatic acquisition of search-guiding heuristics, Proc. CADE 10, Kaiserslautern, FRG, 1990, LNAI 449, pp. 470–484
Slagle, J.R.; Farrell, C.D.: Experiments in automatic learning for a multipurpose heuristic program, Communications of the ACM, Vol. 14, Nr. 2, 1971, pp. 91–99
Slaney, J.:SCOTT: A Model-Guided Theorem Prover, Proc. IJCAI '93, Chambery, FRA, 1993, pp. 109–114
Sutcliffe, G.; Suttner, C.; Yemenis, T.: The TPTP Problem Library, Proc. CADE-12, Nancy, FRA, 1994, LNAI 814, pp. 252–266
Tarski, A.:Logic, Semantics, Metamathematics, Oxford University Press, 1956
Wos, L.: Meeting the Challenge of Fifty Years of Logic, JAR 6, 1990, pp. 213–232
Wos, L.: Searching for Circles of Pure Proofs, JAR 15, 1995, pp. 279–315
Author information
Authors and Affiliations
Editor information
Rights and permissions
Copyright information
© 1996 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Fuchs, M. (1996). Experiments in the heuristic use of past proof experience. In: McRobbie, M.A., Slaney, J.K. (eds) Automated Deduction — Cade-13. CADE 1996. Lecture Notes in Computer Science, vol 1104. Springer, Berlin, Heidelberg. https://doi.org/10.1007/3-540-61511-3_111
Download citation
DOI: https://doi.org/10.1007/3-540-61511-3_111
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-61511-8
Online ISBN: 978-3-540-68687-3
eBook Packages: Springer Book Archive