Latent State Recovery in Reinforcement Learning - John Langford

Explore the intricacies of latent state recovery in reinforcement learning through this comprehensive seminar presented by John Langford from Microsoft Research at the Institute for Advanced Study. Delve into core problems and alternative approaches in reinforcement learning, examining an example problem to understand the challenges involved. Learn about the Homer algorithm, state abstraction techniques, and their proofs. Investigate transition matrices, linear dynamics, and combinatorial state spaces. Engage with a Q&A session and analyze counter-examples to deepen your understanding of this complex topic in theoretical machine learning.

Introduction
Reinforcement Learning
Core Problems
Example Problem
Whats Hard
Questions
Alternative approaches
Hard reinforcement problem
Homer algorithm
State abstraction
Proof
QA
Counter Example
Transition Matrix
Linear Dynamics
combinatorial state space