Title: The Suboptimal Principle of Optimality
Speaker: Mark Chevallier
Abstract: How proving the principle of optimality, within the context of reinforcement learning, has led to me rewriting how I calculate the Q value of a state-action twice in Isabelle – the original proof, intuitive and easy on paper, is much more difficult formally. I will also discuss my current approach to calculating Q values and how it should make proving the Principle of Optimality easier.