Phat C. Vo
← Back to Writing

Course

Deep Reinforcement Learning

A comprehensive course on modern reinforcement learning — from bandits and MDPs to DQN, policy gradients, and model-based methods.

A comprehensive course on modern reinforcement learning.

This series builds Deep RL from the ground up, in four parts:

  • Tabular RL — bandits, Markov decision processes and dynamic programming, Monte Carlo control, temporal-difference learning, function approximation, and the deep-network building blocks.
  • Model-free RL — DQN and its variants, policy gradient, actor-critic (A3C), DDPG, PPO, ACER, and maximum-entropy RL (SAC).
  • Model-based RL — model-based RL, planning with learned world models (PlaNet, Dreamer), importance sampling, MPC, AlphaGo, and natural gradients.
  • Outlook — current research directions: inverse RL, intrinsic motivation, hierarchical, meta, offline, and multi-agent RL.

Read the lectures in order below. Where a slide deck exists, it’s linked at the top of the lecture.

  1. 01 Introduction 10 min
  2. 02 Sampling and Bandits 15 min
  3. 03 Markov Decision Process 23 min
  4. 04 Monte Carlo methods 10 min
  5. 05 Temporal Difference learning 17 min
  6. 06 Function approximation 7 min
  7. 07 Deep learning 12 min
  8. 08 Deep Q-network (DQN) 10 min
  9. 09 DQN variants (Rainbow) 21 min
  10. 10 Policy Gradient methods 13 min
  11. 11 Advantage Actor-Critic (A3C) 14 min
  12. 12 Deep Deterministic Policy Gradient (DDPG) 15 min
  13. 13 Policy optimization (TRPO, PPO) 14 min
  14. 14 Actor-Critic with Experience Replay (ACER) 10 min
  15. 15 Maximum Entropy RL (SAC) 12 min
  16. 16 Model-based RL 4 min
  17. 17 Model-based-augmented model-free RL (Dyna-Q, I2A) 7 min
  18. 18 Off-policy Actor-Critic 20 min
  19. 19 Planning (MPC, TDM) 9 min
  20. 20 World models, Dreamer 8 min
  21. 21 AlphaGo 11 min
  22. 22 Natural gradients 12 min
  23. 23 Intrinsic motivation 1 min