Abductive inference and its applications in program analysis, verification, and synthesis
Işıl Dillig · 2015
Summary form only given. Abductive inference is a form of backwards logical reasoning that infers likely hypotheses from a given conclusion. In other words, given an invalid implication of the form A => B, abduction asks the question "What formula C do we need to conjoin with the antecedent A so that (i) A & C => B is logically valid and (ii) C is consistent with A?" Abductive reasoning has found many applications in program verification and synthesis, particularly in modular program analysis, invariant generation, and automated inference of missing program expressions. This tutorial will give an overview of logical abduction and algorithms for performing abductive inference. We will also survey several use cases of abductive inference in the context of program analysis, verification, and synthesis.