Learning a representation for optimizable formulas
Roughly speaking, a class of formulas K is called optimizable, if there is a class K′, such that for each formula in K there exists a short equivalent formula in K′, where the deduction problem of K′ is solvable in polynomial time. We consider the problem, whether for such classes an optimized representation can be learned using membership, equivalence and clause queries.
Unable to display preview. Download preview PDF.
- 1.H. Aizenstein, L. Hellerstein, L. Pitt: Read-Thrice DNF is Haxd to Learn with Membership and Equivalence Queries, Proc. 31rd Ann. IEEE Symp. on Foundations of Computer Science (1992), 523–532Google Scholar
- 2.D. Angluin: Queries and Concept Learning, Machine Learning 2 (1988), 319–342Google Scholar
- 3.D. Angluin, M. Frazier, L. Pitt: Learning Conjunctions of Horn Clauses, Proc. 31st Ann. IEEE Symp. on Foundations of Computer Science (1990), 186–192Google Scholar
- 5.R. Chang, J. Kadin: On the Structure of Uniquely Satisfiable Formulas, Technical Report TR 90-1124, Dep. of Comp. Sci., Cornell University, Ithaca, N. Y. (1990)Google Scholar
- 7.M. Frazier, L. Pitt: Some New Directions in Computational Learning Theory, Proc. 1st European Conf. on Computational Learning Theory (1993), 19–32Google Scholar
- 9.H. Kleine Büning, U. Löwen: Optimizing Propositional Calculus Formulas with Regard to Questions of Deducibility, Information & Computation 80 (1989), 18–43Google Scholar
- 10.K. Pillaipakkamnatt, V. Raghavan: On the Limits of Proper Learnability of Subclasses of DNF Formulas, Proc. 7th Workshop on Computational Learning Theory (1994), 118–129Google Scholar