On the Parameterized Complexity of Learning Monadic Second-Order Formulas

Within the model-theoretic framework for supervised learning introduced by Grohe and Tur\'an (TOCS 2004), we study the parameterized complexity of learning concepts definable in monadic second-order logic (MSO). We show that the problem of learning a consistent MSO-formula is fixed-parameter tractable on structures of bounded tree-width and on graphs of bounded clique-width in the 1-dimensional case, that is, if the instances are single vertices (and not tuples of vertices). This generalizes previous results on strings and on trees. Moreover, in the agnostic PAC-learning setting, we show that the result also holds in higher dimensions. Finally, via a reduction to the MSO-model-checking problem, we show that learning a consistent MSO-formula is para-NP-hard on general structures.