Query learning of bounded-width OBDDs
In this paper, we study exact learnability of bounded-width ordered binary decision diagrams (OBDDs), when no ordering of the variables is given and learning is conducted by way of equivalence queries and membership queries. We present a learning algorithm for width-2 OBDDs, an algorithm which uses O(n3) equivalence queries alone, where n is the number of variables. We also present a learning algorithm for width-2 OBDDs which uses O(n) proper equivalence queries and O(n2) membership queries. Further, we show a negative result: that there are no polynomial-time algorithms capable of learning width-3 OBDDs from proper equivalence queries alone.
Unable to display preview. Download preview PDF.
- 3.F. Bergadano, N. H. Bshouty, C. Tamon and S. Varricchio. “On Learning Branching Programs and Small Circuits.” Unpublished manuscript.Google Scholar
- 5.N. H. Bshouty, C. Tamon and D. K. Wilson. “On Learning Width Two Branching Programs.” To appear in Proc. of COLT'96.Google Scholar
- 6.F. Ergün, S. Kumar and R. Rubinfeld. “On Learning Bounded-Width Branching Programs.” Proc. of COLT'95, July 1995, pp. 361–368.Google Scholar
- 7.R. Gavaldá and D. Guijarro. “Learning Ordered Binary Decision Diagrams.” Proc. of ALT'95, October 1995, pp. 228–238.Google Scholar
- 8.V. Raghavan and D. Wilkins. “Learning μ-Branching Programs with Queries.” Proc. of COLT'93, July 1993, pp. 27–36.Google Scholar
- 9.H. U. Simon. “Learning Decision Lists and Trees with Equivalence-Queries.” Proc. of EuroCOLT'95, March 1995, pp. 322–336.Google Scholar