A review of Georgia Tech's OMSCS CS7642 Reinforcement Learning course, covering course structure, workload, and key projects. The course spans model-based methods (value/policy iteration), model-free methods (Q-learning, deep Q-learning), and multi-agent reinforcement learning. A highlight project involved training a deep RL agent to land a rocket in OpenAI's LunarLander environment using Double DQN. The course is workload-heavy (30–40 hrs/week), includes three paper-replication projects, weekly homework, and a final exam. Tips include leveraging David Silver's videos and TA office hours.
Table of contents
Why take this course?What’s the course like?Landing the LunarLanderWhat did I learnt?What’s next?1 Impression