Title:
Letting the machines solve your problems in low-dimensional topology
Abstract:
With recent breakthroughs in AI making headlines, it's natural to ask what role machine learning will play in mathematics. In this talk I will outline how machine learning---and reinforcement learning in particular---can be applied to problems in low-dimensional topology. These problems involve finding ribbon disks to rule out potential counterexamples to the smooth 4D Poincare conjecture, constructing minimal length factorizations of braid words, and searching for counterexamples to the Jones unknot conjecture. These applications will range from approaches that have already been successfully implemented to more speculative works-in-progress.