Introduction to Reinforcement Learning 2400-ZEWW967
The course aims to explain the necessary fundamental concepts and algorithms used in Reinforcement Learning, which is a versatile machine learning framework with a wide range of applications.
1.Fundamental terms
a) environment
b) agent
c) reward
d) state
e) action
f) Markov Decision Processes
2. General remarks on Reinforcement Learning
a) Reward Hypothesis
b) Exploration-Exploitation trade-off
c) Challenges in RL
3. Value functions and policies
a) State-value function
b) Action-value function
c) Bellman’s Equation
d) Policy determination
4. Value methods
a) SARSA
b) TD
c) Q learning
d) Deep Q Networks
5. Policy methods
a) REINFORCE
b) DPG
6. Actor-Critic
a) A2C / A3C
b) PPO
c) SAC
7. Basics of Multi-Agent Reinforcement Learning
a) Markov Games
b) Cooperative, competitive and mixed settings
c) Independent and central learning
d) Challenges in MARL
Type of course
Course coordinators
Learning outcomes
Students will understand the core principles of Reinforcement Learning and be able to model decision-making problems as agent-environment interactions. Through hands-on coding exercises, students will learn how to implement and train RL agents in simulated environments, using common tools and frameworks. They will also develop the ability to evaluate and tune agent performance, with an awareness of challenges and issues related to the field. Students will be well equipped to apply Reinforcement Learning techniques to real-world scenarios.
Assessment criteria
The final grade will be determined based on: a home-taken project (80% of the grade) and a project presentation (20% of the grade).
Additional information
Additional information (registration calendar, class conductors, localization and schedules of classes), might be available in the USOSweb system: