Abstract
We describe an efficient implementation of given clause selection in saturation-based automated theorem provers, extending the previous ENIGMA approach. Unlike in the first ENIGMA implementation where a fast linear classifier is trained and used together with manually engineered features, we have started to experiment with more sophisticated state-of-the-art machine learning methods such as gradient boosted trees and recursive neural networks. In particular, the latter approach poses challenges in terms of efficiency of clause evaluation, however, we show that deep integration of the neural evaluation with the ATP data-structures can largely amortize this cost and lead to competitive real-time results. Both methods are evaluated on a large dataset of theorem proving problems and compared with the previous approaches. The resulting methods improve on the manually designed clause guidance, providing the first practically convincing application of gradient-boosted and neural clause guidance in saturation-style automated theorem provers.
Supported by the ERC Consolidator grant no. 649043 AI4REASON, and by the Czech project AI&Reasoning CZ.02.1.01/0.0/0.0/15_003/0000466 and the European Regional Development Fund.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Similar content being viewed by others
Notes
- 1.
- 2.
We use the XGBoost parameter \(\texttt {scale\_pos\_weight}\).
- 3.
Due to various numerical problems with deep recursive networks we have obtained better results with ReLU6, defined by \(\min (\max (0,x),6)\), or \(\tanh \).
- 4.
- 5.
For instance, using the probability (1) for a more fine-grained order on clauses dictated by the neural model.
- 6.
The benchmark can be found at https://github.com/JUrban/MPTP2078. For all the remaining materials for reproducing the experiments please check out the repository https://github.com/ai4reason/eprover-data/tree/master/CADE-19.
- 7.
This appears to be a reasonable waiting time for, e.g., the users of ITP hammers [5].
- 8.
Combining the dimensions for the clause and the conjecture.
- 9.
Moreover, we always try to put examples with the same conjecture G into the same batch to share the time for recomputing the representation of G.
- 10.
For \(\mathcal {M}_\text {lin}\), we show the numbers after five iterations of the boosting loop (see Sect. 3.2). The values in the first round were 40.81% for the positive and 98.62% for the negative rate.
- 11.
We have also measured how much \(\mathcal {S}\) benefits from increased time limits. It solves 1099 problems in 20 s and 1137 problems in 300 s.
- 12.
This metric is similar in spirit to given clause utilization introduced by Schulz and Möhrmann [41].
- 13.
Note that more global caching (of, e.g., whole clauses and frequent combinations of literals) across multiple problems may further amortize the cost of the neural evaluation. This is left as future work here.
References
Alama, J., Heskes, T., Kühlwein, D., Tsivtsivadze, E., Urban, J.: Premise selection for mathematics by corpus analysis and kernel methods. J. Autom. Reason. 52(2), 191–213 (2014)
Alemi, A.A., Chollet, F., Eén, N., Irving, G., Szegedy, C., Urban, J.: DeepMath - deep sequence models for premise selection. In: Lee, D.D., Sugiyama, M., Luxburg, U.V., Guyon, I., Garnett, R. (eds.) Advances in Neural Information Processing Systems 29: Annual Conference on Neural Information Processing Systems 2016, Barcelona, Spain, 5–10 December 2016, pp. 2235–2243 (2016)
Bancerek, G., et al.: Mizar: state-of-the-art and beyond. In: Kerber, M., Carette, J., Kaliszyk, C., Rabe, F., Sorge, V. (eds.) CICM 2015. LNCS (LNAI), vol. 9150, pp. 261–279. Springer, Cham (2015). https://doi.org/10.1007/978-3-319-20615-8_17
Blanchette, J.C., Greenaway, D., Kaliszyk, C., Kühlwein, D., Urban, J.: A learning-based fact selector for Isabelle/HOL. J. Autom. Reason. 57(3), 219–244 (2016)
Blanchette, J.C., Kaliszyk, C., Paulson, L.C., Urban, J.: Hammering towards QED. J. Formalized Reason. 9(1), 101–148 (2016)
Bridge, J.P., Holden, S.B., Paulson, L.C.: Machine learning for first-order theorem proving - learning to select a good heuristic. J. Autom. Reason. 53(2), 141–172 (2014)
Chen, T., Guestrin, C.: XGBoost: a scalable tree boosting system. In: KDD, pp. 785–794. ACM (2016)
Denzinger, J., Fuchs, M., Goller, C., Schulz, S.: Learning from previous proof experience. Technical report AR99-4, Institut für Informatik, Technische Universität München (1999)
Ertel, W., Schumann, J.M.P., Suttner, C.B.: Learning heuristics for a theorem prover using back propagation. In: Retti, J., Leidlmair, K. (eds.) 5. Österreichische Artificial Intelligence-Tagung. INFORMATIK, vol. 208, pp. 87–95. Springer, Heidelberg (1989). https://doi.org/10.1007/978-3-642-74688-8_10
Fan, R.-E., Chang, K.-W., Hsieh, C.-J., Wang, X.-R., Lin, C.-J.: LIBLINEAR: a library for large linear classification. J. Mach. Learn. Res. 9, 1871–1874 (2008)
Färber, M., Brown, C.: Internal guidance for satallax. In: Olivetti and Tiwari [33], pp. 349–361
Färber, M., Kaliszyk, C., Urban, J.: Monte Carlo tableau proof search. In: de Moura, L. (ed.) CADE 2017. LNCS (LNAI), vol. 10395, pp. 563–579. Springer, Cham (2017). https://doi.org/10.1007/978-3-319-63046-5_34
Gauthier, T., Kaliszyk, C.: Premise selection and external provers for HOL4. In: Certified Programs and Proofs (CPP 2015) (2015). https://doi.org/10.1145/2676724.2693173
Goertzel, Z., Jakubův, J., Schulz, S., Urban, J.: ProofWatch: watchlist guidance for large theories in E. In: Avigad, J., Mahboubi, A. (eds.) ITP 2018. LNCS, vol. 10895, pp. 270–288. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-94821-8_16
Goertzel, Z., Jakubuv, J., Urban, J.: ProofWatch meets ENIGMA: first experiments. In: Barthe, G., Korovin, K., Schulz, S., Suda, M., Sutcliffe, G., Veanes, M. (eds.) LPAR-22 Workshop and Short Paper Proceedings. Kalpa Publications in Computing, vol. 9, pp. 15–22. EasyChair (2018)
Goller, C., Küchler, A.: Learning task-dependent distributed representations by backpropagation through structure. In: Proceedings of International Conference on Neural Networks (ICNN 1996), vol. 1, pp. 347–352, June 1996
Gottlob, G., Sutcliffe, G., Voronkov, A. (eds.): Global Conference on Artificial Intelligence, GCAI 2015, Tbilisi, Georgia, 16–19 October 2015. EPiC Series in Computing, vol. 36. EasyChair (2015)
Hochreiter, S., Schmidhuber, J.: Long short-term memory. Neural Comput. 9, 1735–1780 (1997)
Jakubův, J., Urban, J.: ENIGMA: efficient learning-based inference guiding machine. In: Geuvers, H., England, M., Hasan, O., Rabe, F., Teschke, O. (eds.) CICM 2017. LNCS (LNAI), vol. 10383, pp. 292–302. Springer, Cham (2017). https://doi.org/10.1007/978-3-319-62075-6_20
Jakubův, J., Urban, J.: Enhancing ENIGMA given clause guidance. In: Rabe, F., Farmer, W.M., Passmore, G.O., Youssef, A. (eds.) CICM 2018. LNCS (LNAI), vol. 11006, pp. 118–124. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-96812-4_11
Jakubův, J., Urban, J.: Hierarchical invention of theorem proving strategies. AI Commun. 31(3), 237–250 (2018)
Jakubuv, J., Urban, J.: BliStrTune: hierarchical invention of theorem proving strategies. In: Bertot, Y., Vafeiadis, V. (eds.) Proceedings of the 6th ACM SIGPLAN Conference on Certified Programs and Proofs, CPP 2017, Paris, France, 16–17 January 2017, pp. 43–52. ACM (2017)
Joulin, A., Grave, E., Bojanowski, P., Mikolov, T.: Bag of tricks for efficient text classification. In: Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics, Short Papers, vol. 2, pp. 427–431. Association for Computational Linguistics, April 2017
Kaliszyk, C., Urban, J.: Learning-assisted automated reasoning with Flyspeck. J. Autom. Reason. 53(2), 173–213 (2014)
Kaliszyk, C., Urban, J.: FEMaLeCoP: fairly efficient machine learning connection prover. In: Davis, M., Fehnker, A., McIver, A., Voronkov, A. (eds.) LPAR 2015. LNCS, vol. 9450, pp. 88–96. Springer, Heidelberg (2015). https://doi.org/10.1007/978-3-662-48899-7_7
Kaliszyk, C., Urban, J., Michalewski, H., Olsák, M.: Reinforcement learning of theorem proving. In: Bengio, S., Wallach, H.M., Larochelle, H., Grauman, K., Cesa-Bianchi, N., Garnett, R. (eds.) Advances in Neural Information Processing Systems 31: Annual Conference on Neural Information Processing Systems 2018, NeurIPS 2018, Canada, Montréal, 3–8 December 2018, pp. 8836–8847 (2018)
Kaliszyk, C., Urban, J., Vyskocil, J.: Efficient semantic features for automated reasoning over large theories. In: IJCAI, pp. 3084–3090. AAAI Press (2015)
Kovács, L., Voronkov, A.: First-order theorem proving and Vampire. In: Sharygina, N., Veith, H. (eds.) CAV 2013. LNCS, vol. 8044, pp. 1–35. Springer, Heidelberg (2013). https://doi.org/10.1007/978-3-642-39799-8_1
Kuehlwein, D., Urban, J.: Learning from multiple proofs: first experiments. In: Fontaine, P., Schmidt, R.A., Schulz, S. (eds.) PAAR-2012. EPiC Series, vol. 21, pp. 82–94. EasyChair (2013)
Kühlwein, D., van Laarhoven, T., Tsivtsivadze, E., Urban, J., Heskes, T.: Overview and evaluation of premise selection techniques for large theory mathematics. In: Gramlich, B., Miller, D., Sattler, U. (eds.) IJCAR 2012. LNCS (LNAI), vol. 7364, pp. 378–392. Springer, Heidelberg (2012). https://doi.org/10.1007/978-3-642-31365-3_30
Loos, S.M., Irving, G., Szegedy, C., Kaliszyk, C.: Deep network guided proof search. In: Eiter, T., Sands, D. (eds.) LPAR-21, 21st International Conference on Logic for Programming, Artificial Intelligence and Reasoning, Maun, Botswana, 7–12 May 2017. EPiC Series in Computing, vol. 46, pp. 85–105. EasyChair (2017)
Meng, J., Paulson, L.C.: Translating higher-order clauses to first-order clauses. J. Autom. Reason. 40(1), 35–60 (2008)
Olivetti, N., Tiwari, A. (eds.): IJCAR 2016. LNCS (LNAI), vol. 9706. Springer, Cham (2016). https://doi.org/10.1007/978-3-319-40229-1
Otten, J., Bibel, W.: leanCoP: lean connection-based theorem proving. J. Symb. Comput. 36(1–2), 139–161 (2003)
Overbeek, R.A.: A new class of automated theorem-proving algorithms. J. ACM 21(2), 191–200 (1974)
Piotrowski, B., Urban, J.: ATPboost: learning premise selection in binary setting with ATP feedback. In: Galmiche, D., Schulz, S., Sebastiani, R. (eds.) IJCAR 2018. LNCS (LNAI), vol. 10900, pp. 566–574. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-94205-6_37
Polikar, R.: Ensemble based systems in decision making. IEEE Circuits Syst. Mag. 6(3), 21–45 (2006)
Schäfer, S., Schulz, S.: Breeding theorem proving heuristics with genetic algorithms. In: Gottlob et al. [17], pp. 263–274
Schulz, S.: Learning search control knowledge for equational deduction. DISKI, vol. 230. Infix Akademische Verlagsgesellschaft (2000)
Schulz, S.: E - a brainiac theorem prover. AI Commun. 15(2–3), 111–126 (2002)
Schulz, S., Möhrmann, M.: Performance of clause selection heuristics for saturation-based theorem proving. In: Olivetti and Tiwari [33], pp. 330–345
Urban, J.: MPTP 0.2: design, implementation, and initial experiments. J. Autom. Reason. 37(1–2), 21–43 (2006)
Urban, J.: BliStr: the blind strategymaker. In: Gottlob et al. [17], pp. 312–319
Urban, J., Sutcliffe, G., Pudlák, P., Vyskočil, J.: MaLARea SG1 - machine learner for automated reasoning with semantic guidance. In: Armando, A., Baumgartner, P., Dowek, G. (eds.) IJCAR 2008. LNCS (LNAI), vol. 5195, pp. 441–456. Springer, Heidelberg (2008). https://doi.org/10.1007/978-3-540-71070-7_37
Urban, J., Vyskočil, J., Štěpánek, P.: MaLeCoP machine learning connection prover. In: Brünnler, K., Metcalfe, G. (eds.) TABLEAUX 2011. LNCS (LNAI), vol. 6793, pp. 263–277. Springer, Heidelberg (2011). https://doi.org/10.1007/978-3-642-22119-4_21
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2019 Springer Nature Switzerland AG
About this paper
Cite this paper
Chvalovský, K., Jakubův, J., Suda, M., Urban, J. (2019). ENIGMA-NG: Efficient Neural and Gradient-Boosted Inference Guidance for E. In: Fontaine, P. (eds) Automated Deduction – CADE 27. CADE 2019. Lecture Notes in Computer Science(), vol 11716. Springer, Cham. https://doi.org/10.1007/978-3-030-29436-6_12
Download citation
DOI: https://doi.org/10.1007/978-3-030-29436-6_12
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-29435-9
Online ISBN: 978-3-030-29436-6
eBook Packages: Computer ScienceComputer Science (R0)