Skip to main content

Reverse Engineering and Learning of Temporal Logic Queries

Venue: Birkbeck Main Building, Malet Street

In reverse engineering of database queries, one aims to learn a query from given sets of examples where answers are suposed to be positive and negative. The query can then be used to find explanation of the answers and non-answers, and also as a classifier for new data instances. We start by formulating a general separation problem in the setting where the queries are temporal logic formulas, and the data instances are finite sets of temporal facts. We present a number of examples to motivate the problem for such logics as LTL, metric temporal logic, and interval temporal logic. For LTL, we present our recently obtained results related to the computational complexity of computing separating queries.

Contact name: