Overview
Explore the statistical complexity of reinforcement learning in this 30-minute lecture by Sham Kakade from Harvard University, presented at the Simons Institute 10th Anniversary Symposium. Delve into key concepts including supervised learning, notation, low bounds, and challenges in the field. Examine the linear model, evaluation problem, and online problem, while considering linear realizability and sufficient conditions. Investigate the completeness assumption and special cases before concluding with a comprehensive wrap-up of the topic.
Syllabus
Intro
Supervised Learning
Notation
Low Bounds
Challenges
Linear Model
Evaluation Problem
Online Problem
Linear Realizability
Sufficient Conditions
completeness assumption
special cases
wrap up
Taught by
Simons Institute