Distributional RL

Overview

Explore distributional reinforcement learning in this 23-minute video lecture from Pascal Poupart's CS885 course at the University of Waterloo. Delve into key concepts including return distribution, policy evaluation, convergence, and the Bellman equation. Examine the C51 (Categorical DQN) algorithm, its advantages, and its performance on Atari games. Gain insights into various distributional representations and their applications in reinforcement learning. Access accompanying slides from the course website to enhance your understanding of this advanced topic in machine learning and artificial intelligence.

Syllabus

Outline
Objective
Distributional RL
Return Distribution
Policy Evaluation
Convergence
Bellman Equation
C51 (Categorical DQN)
Advantage
Atari Results
Distributional Representations

Taught by

Pascal Poupart

Reviews

Start your review of Distributional RL

Taught by

DQN - Playing Atari with Deep Reinforcement Learning - RL Paper Explained

Maximum Entropy Reinforcement Learning

Inverse RL

Towards Structural Risk Minimization for RL - Emma Brunskill

Deep Robust Reinforcement Learning and Regularization

On the Statistical Complexity of Reinforcement Learning

Never Stop Learning.