Learning-powered computer-assisted counterexample search

Abstract

[en] This thesis explores the great potential of computer-assisted proofs in the advancement of mathematical knowledge, with a special focus on using computers to refute conjectures by finding counterexamples, sometimes a humanly impossible task. In recent years, mathematicians have become more aware that machine learning techniques can be extremely helpful for finding counterexamples to conjectures in a more efficient way than by using exhaustive search methods. In this thesis we do not only present the theoretical background behind some of these methods but also implement them to try to refute some graph theory conjectures.