CENG045 Reinforcement Learning

Information

Course Goal

The goal of a reinforcement learning course is to teach students the fundamentals of reinforcement learning, which is a subfield of machine learning. Reinforcement learning is concerned with how agents can learn to make decisions in an environment to achieve a specific goal.

Course Content

This course covers the Introduction to Reinforcement Learning, Basic concepts of reinforcement learning, comparison with supervised and unsupervised learning, and types of reinforcement learning problems, Markov Decision Processes (MDPs), Formalism of MDPs, reward function, state transitions, policy, value function, and Bellman equations, Dynamic Programming (DP): Policy evaluation, policy iteration, value iteration, and Monte Carlo methods. Temporal Difference (TD) Learning: On-policy and off-policy learning, Q-learning, SARSA, and eligibility traces. Function Approximation: Linear and non-linear function approximation, and deep reinforcement learning. Exploration and Exploitation: Exploration strategies such as epsilon-greedy, softmax, and UCB.

Course Precondition

Knowledge of basic programming, linear algebra, and probability theory.

Resources

Sutton, Richard S., and Andrew G. Barto. Reinforcement learning: An introduction. MIT press, 2018.

Notes

Sutton, Richard S., and Andrew G. Barto. Reinforcement learning: An introduction. MIT press, 2018.

Course Learning Outcomes

Relation with Program Learning Outcome

Week Plan

Student Workload - ECTS