Rigorous Results from Machine Learning Using Reinforcement Learning

Watch a 32-minute lecture from the CMSA Mathematics and Machine Learning Closing Workshop where Fabian Ruehle from Northeastern University explores the application of Reinforcement Learning (RL) in mathematical proofs. Discover how RL techniques can be leveraged to obtain provably correct mathematical results, beginning with a foundational introduction to Reinforcement Learning principles. Follow along as Ruehle demonstrates practical applications through three distinct examples: a Number Theory case involving Diophantine Equations connected to String Theory, and two Knot Theory problems - one focused on unknot identification and another examining ribbon knots, which are particularly relevant to investigations of the smooth Poincare conjecture.