Perquisites: MET CS 677 or instructor’s consent. - This course focuses on reinforcement learning, covering fundamental concepts and advanced techniques. It begins with an introduction to reinforcement learning and key concepts, such as exploitation versus exploration and Markov Decision Processes. As the course progresses, it delves into state transition diagrams, the Bellman equation, and solutions to the Multi-Armed Bandits problem. Students will explore challenges and methods related to control and prediction. Then, they learn tabular methods, including Monte Carlo, Dynamic Programming, Temporal Difference Learning, SARSA, and Q-Learning. Afterwards, the course also extends into reviewing neural network concepts, covering convolutional and recurrent neural networks, and moves on to approximation methods for both discrete and continuous spaces, including DQN and its variants. Policy gradient methods, actor-critic methods. Finally, ethical considerations in AI and safety issues are also discussed.

Note that this information may change at any time. Please visit the MyBU Student Portal for the most up-to-date course information.