Pulina, Luca and Tacchella, Armando (2009) Learning to integrate deduction and search in reasoning about quantified boolean formulas. In: FroCoS'09: 7th international conference on Frontiers of combining systems: proceedings, 16-18 September 2009, Trento, Italy. Berlin - Heidelberg, Springer. p. 350-365. (Lecture Notes in Computer Science, 5749/2009). ISBN 978-3-642-04221-8. Conference or Workshop Item.
Full text not available from this repository.
In this paper we study the problem of integrating deduction and search with the aid of machine learning techniques to yield practically efficient decision procedures for quantified Boolean formulas (QBFs). We show that effective on-line policies can be learned from the observed performances of deduction and search on representative sets of formulas. Such policies can be leveraged to switch between deduction and search during the solving process. We provide empirical evidence that learned policies perform better than either deduction and search, even when the latter are combined using hand-made policies based on previous works. The fact that even with a proof-of-concept implementation, our approach is competitive with sophisticated state-of-the-art QBF solvers shows the potential of machine learning techniques in the integration of different reasoning methods.
I documenti depositati in UnissResearch sono protetti dalle leggi che regolano il diritto d'autore
Repository Staff Only: item control page