Necessary and sufficient conditions for learning with correction queries

Abstract

We investigate the newly introduced model of learning with correction queries in the context of query learning. We present necessary and sufficient conditions for a class of languages to be inferable within this setting. We also offer a complete picture of how is the model of learning with corrections related with other well-established learning models, like the model of learning in the limit from positive data, or the one of learning with membership queries. As an application, we show that the class of k -reversible languages is learnable with correction queries.