Widening and Interpolation

Kenneth L. McMillan

Abstract: Widening/narrowing and interpolation are two techniques for deriving a generalization about unbounded behaviors from an analysis of bounded behaviors. The purpose of both methods is to produce an inductive invariant that proves some property of a program or other discrete dynamic system. In the case of widening, we obtain a guess at an inductive invariant by extrapolating a sequence of approximations of the program behavior, either forward or backward. In the case of interpolation, we use the intermediate assertions in proofs of correctness of bounded behaviors.

External IDs:dblp:conf/sas/McMillan11