Property Testing for Bounded Degree Databases

Aiming at extremely efficient algorithms for big data sets, we introduce property testing of relational databases of bounded degree. Our model generalises the bounded degree model for graphs (Goldreich and Ron, STOC 1997). We prove that in this model, if the databases have bounded tree-width, then every query definable in monadic second-order logic with modulo counting is testable with a constant number of oracle queries and polylogarithmic running time. This is the first logical meta-theorem in property testing of sparse models. Furthermore, we discuss conditions for the existence of uniform and non-uniform testers.