Аннотация:
Automating the search for proofs by induction is an imporant topic in computer
science with a history that stretches back decades. A variety of different
approaches and systems has been developed. Typically, these systems have been
evaluated empirically and thus very little is known about their theoretical
limitations.
In this talk I will show how to use mathematical logic to understand the
theoretical power and limits of methods for automated inductive theorem
proving. A central tool are translations of proof systems that are intended for
automated proof search into (very) weak arithmetical theories. Thus
unprovability results can be transferred from theories to methods of automated
deduction.
I will describe concrete such analyses of two methods of automated inductive
theorem proving including practically relevant unprovability results: 1. adding
explicit induction axioms to a saturation theorem prover and 2. clause set
cycles.
(Joint work with Jannik Vierling.)