Title of the course: Reinforcement Learning
Instructor: Dr. Naci Saldı
Institution: Bilkent Üniversitesi
Dates: 19-25 August 2024
Prerequisites: Linear algebra, probability theory, basic real analysis
Level: Graduate, advanced undergraduate
Abstract: In this course, we aim to introduce the concept of reinforcement learning. We’ll start by giving an overview of Markov decision processes, defining the model, and explaining the dynamic programming principle. Then, we’ll discuss the value iteration algorithm, which helps find the optimal policy and value function when the model is known. For cases where the model is unknown, we’ll delve into reinforcement learning methods. Specifically, we’ll focus on Q-learning, a well-known algorithm used for dealing with uncertain dynamics and costs. Additionally, we’ll examine the proof of convergence for the Q-learning algorithm using stochastic approximation techniques. By the end of the course, you’ll have a solid understanding of reinforcement learning.
Language: English