Reinforcement Learning: Hidden Theory and New Super-Fast Algorithms

Explore cutting-edge advancements in reinforcement learning through this 1-hour 6-minute lecture by Sean Meyn from the University of Florida. Delve into the hidden theory behind reinforcement learning and discover new super-fast algorithms that are revolutionizing the field. Begin with an introduction and overview, followed by a recap of fundamental concepts. Examine the intricacies of hidden theory, including choleric and relaxation techniques, and their application to finite state spaces. Investigate Watkins' key function and its role in conditional expectation. Learn about the Watkins algorithm and its limitations, leading to the introduction of matrix-based approaches and one-to-one mappings. Apply these concepts to real-world scenarios such as local conversion, trading stocks, and finance. Conclude by discussing future work and potential applications in the rapidly evolving domain of reinforcement learning.

Introduction
Overview
Recap
Hidden Theory
choleric and relaxation
finite state space
Watkins key function
Conditional expectation
Watkins algorithm
We need help
A matrix
Onetoone mapping
Local Conversion
Trading Stocks
Finance
Future Work