Detection of knots in polymers with machine learning

Knotting is a fascinating phenomenon naturally occurring in chain-like molecules such as polymers, DNA, and proteins, and can influence their properties and dynamics. The classification of knot equivalences is one of the central problems in knot theory, and work in this field led to the development of topological invariants to distinguish knot types. However, different knots can share the same invariants, and their calculation can be computationally expensive, especially when long chains or highly intricate polymer conformations are considered. The aim of the present thesis is to provide a detailed and comprehensive review of the major developments that have been made over the past few years in the use of artificial neural networks as a promising computational method for knot identification in polymers and proteins, and to present the state of the art in this interdisciplinary and evolving research field.