Using Machine Learning to Formulate Mathematical Conjectures - IPAM at UCLA

Explore how machine learning can be utilized to formulate mathematical conjectures in this insightful lecture by Marc Lackenby from the University of Oxford. Delve into the application of supervised learning in discovering connections between different areas of low-dimensional topology and geometry, focusing on predicting knot signatures from hyperbolic invariants. Examine the process of formulating and proving precise conjectures based on machine learning insights, while also discussing the method's limitations, including challenges in interpreting patterns and addressing outliers. Gain valuable insights into the potential applications of this approach across various mathematical fields and discover new examples of unexpected conjectural connections in low-dimensional topology uncovered through machine learning techniques.

Intro
Supervised learning
The branches of knot theory
Connections with dimension 4
Using machine learning in knot theory
Hyperbolic structures
Signature and cusp geometry
The natural slope
First conjectures
Difficulties with this method