Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
126 tokens/sec
GPT-4o
47 tokens/sec
Gemini 2.5 Pro Pro
43 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
47 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Efficiently Explaining CSPs with Unsatisfiable Subset Optimization (extended algorithms and examples) (2303.11712v3)

Published 21 Mar 2023 in cs.AI and cs.LO

Abstract: We build on a recently proposed method for stepwise explaining solutions of Constraint Satisfaction Problems (CSP) in a human-understandable way. An explanation here is a sequence of simple inference steps where simplicity is quantified using a cost function. The algorithms for explanation generation rely on extracting Minimal Unsatisfiable Subsets (MUS) of a derived unsatisfiable formula, exploiting a one-to-one correspondence between so-called non-redundant explanations and MUSs. However, MUS extraction algorithms do not provide any guarantee of subset minimality or optimality with respect to a given cost function. Therefore, we build on these formal foundations and tackle the main points of improvement, namely how to generate explanations efficiently that are provably optimal (with respect to the given cost metric). For that, we developed (1) a hitting set-based algorithm for finding the optimal constrained unsatisfiable subsets; (2) a method for re-using relevant information over multiple algorithm calls; and (3) methods exploiting domain-specific information to speed up the explanation sequence generation. We experimentally validated our algorithms on a large number of CSP problems. We found that our algorithms outperform the MUS approach in terms of explanation quality and computational time (on average up to 56 % faster than a standard MUS approach).

Definition Search Book Streamline Icon: https://streamlinehq.com
References (55)
  1. Finding good proofs for description logic entailments using recursive quality measures.  Automated Deduction-CADE, 12–15.
  2. Description logics.  In Handbook on ontologies, pp. 3–28. Springer.
  3. Using minimal correction sets to more efficiently compute minimal unsatisfiable sets.  In Computer Aided Verification: 27th International Conference, CAV 2015, San Francisco, CA, USA, July 18-24, 2015, Proceedings, Part II, pp. 70–86. Springer.
  4. Finding a collection of muses incrementally.  In Integration of AI and OR Techniques in Constraint Programming: 13th International Conference, CPAIOR 2016, Banff, AB, Canada, May 29-June 1, 2016, Proceedings 13, pp. 35–44. Springer.
  5. Must: minimal unsatisfiable subsets enumeration tool.  In Tools and Algorithms for the Construction and Analysis of Systems: 26th International Conference, TACAS 2020, Held as Part of the European Joint Conferences on Theory and Practice of Software, ETAPS 2020, Dublin, Ireland, April 25–30, 2020, Proceedings, Part I 26, pp. 135–152. Springer.
  6. Replication-guided enumeration of minimal unsatisfiable subsets.  In Principles and Practice of Constraint Programming: 26th International Conference, CP 2020, Louvain-la-Neuve, Belgium, September 7–11, 2020, Proceedings 26, pp. 37–54. Springer.
  7. Handbook of Satisfiability.
  8. Step-wise explanations of constraint satisfaction problems.  In Proceedigns of ECAI, pp. 640–647.
  9. A framework for step-wise explaining how to solve constraint satisfaction problems.  Artificial Intelligence, 300, 103550.
  10. Plan explanations as model reconciliation: moving beyond explanation as soliloquy.  In Proceedings of IJCAI, pp. 156–163.
  11. Argumentation for explainable scheduling.  In Proceedings of AAAI, pp. 2752–2759.
  12. Exploiting the power of MIP solvers in MAXsat.  In Proceedings of SAT, pp. 166–181.
  13. A scalable algorithm for minimal unsatisfiable core extraction.  In Proceedings of SAT, pp. 36–41.
  14. Using small muses to explain how to solve pen and paper puzzles.  ArXiv, abs/2104.15040.
  15. FET (2019). Fetproact-eic-05-2019, fet proactive: emerging paradigms and communities, call.. Horizon 2020 Framework Programme.
  16. Explainable planning.  In Proceedings of IJCAI’17-XAI.
  17. Explanation and implication for configuration problems.  In IJCAI 2001 workshop on configuration, pp. 31–37.
  18. Essence: A constraint language for specifying combinatorial problems.  Constraints, 13(3), 268–306.
  19. Efficiently explaining CSPs with unsatisfiable subset optimization.  In Zhou, Z.-H. (Ed.), Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence, IJCAI-21, pp. 1381–1388. International Joint Conferences on Artificial Intelligence Organization. Main Track.
  20. An approach for extracting a small unsatisfiable core.  Formal Methods in System Design, 33(1-3), 1–27.
  21. Verification of proofs of unsatisfiability for CNF formulas.  In Proceedings of DATE, pp. 10886–10891.
  22. A survey of methods for explaining black box models.  ACM computing surveys (CSUR), 51(5), 1–42.
  23. Gunning, D. (2017). Explainable artificial intelligence (XAI).  Defense Advanced Research Projects Agency, 2.
  24. Robustness and explainability of artificial intelligence.  Publications Office of the European Union.
  25. Proceedings of FAT*.
  26. Huang, J. (2005). Mup: A minimal unsatisfiability prover.  In Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC, pp. 432– 437 Vol. 1.
  27. Quantified maximum satisfiability.  Constraints, 21(2), 277–302.
  28. PySAT: A Python toolkit for prototyping with SAT oracles.  In SAT, pp. 428–437.
  29. Abduction-based explanations for machine learning models.  In Proceedings of AAAI, pp. 1511–1519.
  30. Smallest MUS extraction with minimal hitting set dualization.  In Proceedings of CP.
  31. Junker, U. (2001). QuickXPlain: Conflict detection for arbitrary constraint propagation algorithms.  In IJCAI’01 Workshop on Modelling and Solving problems with constraints.
  32. Theory of quantified boolean formulas.  In Handbook of Satisfiability, pp. 735–760.
  33. Koopmann, P. (2021). Two ways of explaining negative entailments in description logics using abduction.  Explainable Logic-Based Knowledge Representation (XLoKR 2021).
  34. Explainable agency for intelligent autonomous systems.  In Twenty-Ninth IAAI Conference.
  35. MaxSAT, hard and soft constraints.  In Handbook of satisfiability, pp. 903–927. IOS Press.
  36. Explanation semantics for abstract argumentation.  In Computational Models of Argument, pp. 271–282. IOS Press.
  37. Fast, flexible mus enumeration.  Constraints, 21(2), 223–250.
  38. Algorithms for computing minimal unsatisfiable subsets of constraints.  J. Autom. Reasoning, 40(1), 1–33.
  39. A unified approach to interpreting model predictions.  In Proceedings of NIPS, pp. 4765–4774.
  40. On computing minimum unsatisfiable cores.  In Proceedings of SAT.
  41. Marques-Silva, J. (2010). Minimal unsatisfiability: Models, algorithms and applications.  In 2010 40th IEEE International Symposium on Multiple-Valued Logic.
  42. On computing minimal correction subsets.  In Twenty-Third International Joint Conference on Artificial Intelligence.
  43. Miller, T. (2019). Explanation in artificial intelligence: Insights from the social sciences.  Artificial Intelligence, 267, 1–38.
  44. Proceedings of the IJCAI 2019 Workshop on Explainable Artificial Intelligence.
  45. The added value of argumentation.  In Agreement technologies, pp. 357–403. Springer.
  46. Automatically improving constraint models in savile row.  Artificial Intelligence, 251, 35–61.
  47. Using description logic and abox abduction to capture medical diagnosis.  In International Conference on Industrial, Engineering and Other Applications of Applied Intelligent Systems, pp. 376–388. Springer.
  48. AMUSE: a minimally-unsatisfiable subformula extractor.  In Proceedings of DAC, pp. 518–523.
  49. Reiter, R. (1987). A theory of diagnosis from first principles.  AIJ, 32(1), 57–95.
  50. Handbook of Constraint Programming, Vol. 2 of Foundations of Artificial Intelligence. Elsevier.
  51. Implicit hitting set algorithms for reasoning beyond NP.  In Proceedings of KR, pp. 104–113.
  52. Abstract argumentation and explanation applied to scientific debates.  Synthese, 190(12), 2195–2217.
  53. Inference-based constraint satisfaction supports explanation.  In AAAI/IAAI, Vol. 1, pp. 318–325.
  54. Strong explanations in abstract argumentation.  In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 35, pp. 6496–6504.
  55. Argumentation and explainable artificial intelligence: a survey.  The Knowledge Engineering Review, 36.
Citations (6)
View on Semantic Scholar

Summary

We haven't generated a summary for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Now