Abstract
A major problem in computational learning theory is whether the class of formulas in conjunctive normal form (CNF) is efficiently learnable. Although it is known that this class cannot be polynomially learned using either membership or equivalence queries alone, it is open whether the CNF class can be polynomially learned using both types of queries. One of the most important results concerning a restriction of the CNF class is that propositional Horn formulas are polynomial time learnable in Angluin’s exact learning model with membership and equivalence queries. In this work, we push this boundary and show that the class of multivalued dependency formulas (MVDF), which non-trivially extends propositional Horn, is polynomially learnable from interpretations. We then provide a notion of reduction between learning problems in Angluin’s model, showing that a transformation of the algorithm suffices to efficiently learn multivalued database dependencies from data relations. We also show via reductions that our main result extends well known previous results and allows us to find alternative solutions for them.
- Dana Angluin. 1987. Learning k-term DNF Formulas Using Queries and Counterexamples. Technical Report. Department of Computer Science, Yale University.Google Scholar
- Dana Angluin. 1987. Learning regular sets from queries and counterexamples. Inf. Comput. 75, 2 (1987), 87--106.Google Scholar
Digital Library
- Dana Angluin. 1988. Queries and concept learning. Mach. Learn. 2, 4 (1988), 319--342.Google Scholar
Digital Library
- Dana Angluin. 1990. Negative results for equivalence queries. Mach. Learn. 5 (1990), 121--150.Google Scholar
Digital Library
- Dana Angluin, Michael Frazier, and Leonard Pitt. 1992. Learning conjunctions of Horn clauses. Mach. Learn. 9 (1992), 147--164.Google Scholar
Digital Library
- Dana Angluin and Michael Kharitonov. 1995. When won't membership queries help? J. Comput. Syst. Sci. 50, 2 (1995), 336--355.Google Scholar
Digital Library
- Marta Arias and José L. Balcázar. 2011. Construction and learnability of canonical Horn formulas. Mach. Learn. 85, 3 (2011), 273--297.Google Scholar
Digital Library
- José L. Balcázar and Jaume Baixeries. 2004. Characterizations of multivalued dependencies and related expressions. In Proceedings of the 7th International Conference on Discovery Science (DS’04). 306--313.Google Scholar
- N. H. Bshouty. 1995. Exact learning boolean functions via the monotone theory. Inf. Comput. 123, 1 (1995), 146--153.Google Scholar
Digital Library
- Edgar F. Codd. 1970. A relational model of data for large shared data banks. Commun. ACM 13, 6 (1970), 377--387.Google Scholar
Cross Ref
- Ronald Fagin. 1977. Functional dependencies in a relational database and propositional logic. IBM J. Res. Dev. 21, 6 (1977), 534--544.Google Scholar
Digital Library
- Ronald Fagin. 1977. Multivalued dependencies and a new normal form for relational databases. ACM Trans. Database Syst. 2 (1977), 262--278.Google Scholar
Digital Library
- Peter A. Flach and Iztok Savnik. 1999. Database dependency discovery: A machine learning approach. AI Commun. 12, 3 (1999), 139--160.Google Scholar
Digital Library
- Michael Frazier and Leonard Pitt. 1993. Learning from entailment: An application to propositional Horn sentences. In Proceedings of the 10th International Conference on Machine Learning. 120--127.Google Scholar
Cross Ref
- Michael Duane Frazier. 1994. Matters Horn and Other Features in the Computational Learning Theory Landscape: The Notion of Membership. Ph.D. Dissertation. University of Illinois Urbana. 36--37.Google Scholar
- Lisa Hellerstein and Vijay Raghavan. 2002. Exact learning of DNF formulas using DNF hypotheses. In Proceedings of the 34th Annual ACM Symposium on Theory of Computing (STOC’02). ACM, New York, NY, 465--473.Google Scholar
Digital Library
- Montserrat Hermo and Ana Ozaki. 2015. Exact learning of multivalued dependencies. In Proceedings of the 26th International Conference on Algorithmic Learning Theory. Springer, 73--88.Google Scholar
Digital Library
- Montserrat Hermo and Ana Ozaki. 2018. Exact learning of multivalued dependency formulas. Theor. Comput. Sci. 716, 15 (2018), 4--14. DOI:https://doi.org/10.1016/j.tcs.2017.11.018Google Scholar
Cross Ref
- Ykä Huhtala, Juha Kärkkäinen, Pasi Porkka, and Hannu Toivonen. 1998. Efficient discovery of functional and approximate dependencies using partitions. In Proceedings of the 14th International Conference on Data Engineering 1998. IEEE, 392--401.Google Scholar
Cross Ref
- Martti Kantola, Heikki Mannila, Kari-Jouko Räihä, and Harri Siirtola. 1992. Discovering functional and inclusion dependencies in relational databases. Int. J. Intell. Syst. 7, 7 (1992), 591--607.Google Scholar
Cross Ref
- Michael Kearns, Ming Li, Leonard Pitt, and Leslie Valiant. 1987. On the learnability of boolean formulae. In Proceedings of the 19th Annual ACM Symposium on Theory of Computing. 285--295.Google Scholar
Digital Library
- Michael J. Kearns and Umesh Vazirani. 1994. An Introduction to Computational Learning Theory. MIT Press. 155--156.Google Scholar
- Boris Konev, Ana Ozaki, and Frank Wolter. 2016. A model for learning description logic ontologies based on exact learning. In Proceedings of the Conference on Artificial Intelligence (AAAI’16). AAAI Press, 1008--1015.Google Scholar
- Víctor Lavín. 2011. On learning multivalued dependencies with queries. Theor. Comput. Sci. 412, 22 (2011), 2331--2339.Google Scholar
Digital Library
- Víctor Lavín. 2015. Learning an extension of the class of functional dependencies with queries. New Generat. Comput. 33, 3 (2015), 319--340.Google Scholar
Cross Ref
- Víctor Lavín and Montserrat Hermo. 2011. Negative results on learning multivalued dependencies with queries. Inform. Process. Lett. 111, 19 (2011), 968--972.Google Scholar
Digital Library
- Nick Littlestone. 1988. Learning quickly when irrelevant attributes abound: A new linear-threshold algorithm. Mach. Learn. 2, 4 (April 1988), 285--318.Google Scholar
Cross Ref
- Heikki Mannila and Kari-Jouko Räihä. 1994. Algorithms for inferring functional dependencies from relations. Data Knowl. Eng. 12, 1 (1994), 83--99.Google Scholar
Digital Library
- Krishnan Pillaipakkamnatt and Vijay Raghavan. 1995. Read-twice DNF formulas are properly learnable. Information and Computation 122, 2 (1995), 236--267.Google Scholar
Digital Library
- Leonard Pitt and Manfred K. Warmuth. 1988. Reductions among prediction problems: On the difficulty of predicting automata. In Proceedings of the 3rd Annual Structure in Complexity Theory Conference, 1988. 60--69.Google Scholar
- Víctor Lavín Puente. 2015. Learning sets of antecedent-restricted functional and multivalued dependencies with queries. Theory Comput. Syst. (2015), 1--17.Google Scholar
- Yehoshua Sagiv, Claude Delobel, D. Scott Parker, Jr., and Ronald Fagin. 1981. An equivalence between relational database dependencies and a fragment of propositional logic. J. ACM 28, 3 (1981), 435--453.Google Scholar
Digital Library
- Leslie G. Valiant. 1984. A theory of the learnable. Commun. ACM 27, 11 (1984), 1134--1142.Google Scholar
Digital Library
- Osamu Watanabe. 1994. A framework for polynomial-time query learnability. Theory Comput. Syst. 27, 3 (1994), 211--229.Google Scholar
Digital Library
Index Terms
Exact Learning: On the Boundary between Horn and CNF
Recommendations
Exact learning of read-twice DNF formulas (extended abstract)
SFCS '91: Proceedings of the 32nd annual symposium on Foundations of computer scienceA polynomial-time algorithm is presented for exactly learning the class of read-twice DNF formulas, i.e. Boolean formulas in disjunctive normal form where each variable appears at most twice. The (standard) protocol used allows the learning algorithm to ...
Exact learning via the Monotone theory
SFCS '93: Proceedings of the 1993 IEEE 34th Annual Foundations of Computer ScienceWe study the learnability of concept classes from membership and equivalence queries. We develop the Monotone theory that proves (1) Any boolean function is learnable as decision tree. (2) Any boolean function is either learnable as DNF or as CNF (or ...
Efficient learning algorithms yield circuit lower bounds
We describe a new approach for understanding the difficulty of designing efficient learning algorithms. We prove that the existence of an efficient learning algorithm for a circuit class C in Angluin's model of exact learning from membership and ...






Comments