Gauss realizability is not so simple and can confuse both humans and AI

12^th May 2023, 11:00
Alexei Lisitsa

Abstract

Gauss realizability is a concept in topology used in studying planar curves and knots. We apply an experimental approach using constraint satisfaction and logic programming for verification of some recently published simple Gauss realizability criteria and show they are incorrect. We then report on applications of machine learning to the problem of Gauss realizability and demonstrate that ML "is confused'' on the same instances which demonstrated incorrectness above.