Exact learning of multivalued dependency formulas

The transformation of a relational database schema into fourth normal form, which minimizes data redundancy, relies on the correct identification of multivalued dependencies. In this work, we study the learnability of multivalued dependency formulas (MVDF), which correspond to the logical theory behind multivalued dependencies. As we explain, MVDF lies between propositional Horn and 2-Quasi-Horn. We prove that MVDF is polynomially learnable in Angluin et al.'s exact learning model with membership and equivalence queries, provided that counterexamples and membership queries are formulated as 2-Quasi-Horn clauses. As a consequence, we obtain that the subclass of 2-Quasi-Horn theories which are equivalent to MVDF is polynomially learnable.