This course on Reinforcement Learning aims to teach learners the basic theory behind RL and how to train an AI agent using Python libraries like Farama Foundation Gymnasium and Stable Baselines3. The course covers topics such as the environment-agent interaction loop, Markov decision processes, Bellman equations, exploration vs. exploitation, and modern RL algorithms like Q-learning and Deep Q-Networks. By the end of the course, learners will be able to apply their knowledge to solve control theory problems like the classic cartpole and inverted pendulum. The course is suitable for individuals interested in machine learning, AI, and control theory.
Overview
Syllabus
- Intro
- History of reinforcement learning
- Environment and agent interaction loop
- Gymnasium and Stable Baselines3
- Hands-on: how to set up a gymnasium environment
- Markov decision process
- Bellman equation for the state-value function
- Bellman equation for the action-value function
- Bellman optimality equations
- Exploration vs. exploitation
- Recommended textbook
- Model-based vs. model-free algorithms
- On-policy vs. off-policy algorithms
- Discrete vs. continuous action space
- Discrete vs. continuous observation space
- Overview of modern reinforcement learning algorithms
- Q-learning
- Deep Q-network DQN
- Hands-on: how to train a DQN agent
- Usefulness of reinforcement learning
- Challenge: inverted pendulum
- Conclusion
Taught by
Digi-Key
