Ens'IA : Reinforcement Learning course

Introduction

As a member of Ens’IA, I give lectures and lab works about artificial intelligence.

In 2021/2022, @Hugo Cartigny and I gave lectures and lab works about

• Machine Learning with KNN and K-Means
• Deep learning in computer vision (perceptron, sigmoid neuron, neural networks, convolutional neural nerworks …)
• Data processing for computer vision

As you can see, we did not talk about Reinforcement Learning yet, so I dicided to build a course about it, especially about Q-Learning

What is Reinforcement Learning

Reinforcement Learning (RL) is a field of Machine Learning (ML) based on the idea of reward. An agent (a bot in a game for example) is taking actions (moving for instance) in an environment (e.g. the map). As a result, the agent get a reward and is a new state in the environment. Most of the time, the agent goal is to maximize the reward. To do so, the agent take actions and learn through trials and errors.

Q-Learning in a few words

Q-Learning is a way of learning the “value” of a stage. It represents the score the agent can expect to get by going into a particular state

More details and labwork

Lecture slides and lab work with a correction are available under MIT licence