IIT Madras RL Lectures: Reinforcement Learning
BSDA500744 lectures10 weeks
Weeks
- Week 16 lecturesIntroduction to RL · Tic-tac-toe and TD learning · Immediate RL and bandits · Regret and (probably Approximately Correct) frameworks · Upper Confidence Bound (UCB) Algorithm · Contextual bandits
- Week 27 lecturesFull RL problem · Markov decision process (MDP) · Markov decision process (MDP): problem to formulation · Returns · Value functions · Markov decision process (MDP) example: recycling robot · Markov decision process (MDP) example: haunted house
- Week 34 lecturesBellman equations · Bellman optimality equations · Dynamic programming (DP), poilcy iteration (policy evaluation) · Dynamic programming (DP): policy iteration (policy improvement)
- Week 44 lecturesDynamic programming (DP): value iteration · More on dynamic programming (DP) · Monte carlo methods · Comparison of dynamic programming (DP), MC and TD
- Week 55 lecturesTemporal difference learning (TD) · MC vs TD updates · SARSA · Q-learning · Off-policy learning
- Week 62 lecturesN-step prediction and TD-lambda · Double Q learning
- Week 76 lecturesIntroduction to function approximation · Linear function approximation · State and action representation · More on representations · DQN: deep Q-network · Advanced value based methods
- Week 83 lecturesPolicy gradient algorithms · Reinforce · Policy gradient theorem
- Week 104 lecturesReinforce: MC policy gradient · Actor critic methods | part 1 · Actor critic methods | part 2 · Deterministic policy gradient (DPG) and deep deterministic policy gradient (DDPG)
- Week 121 lectureMonte carlo tree search
More lectures
Video 1 · 45:25