REINFORCEMENT LEARNING

Learning Outcome

Understand and learn the key elements in Reinforcement Learning’s theory

Understand and implement the different Reinforcement Learning algorithms for prediction and control (model-based and model-free)

Being able to implement a simple RL problem in Python

Having an overview of the use-cases of Reinforcement Learning and knowing when to use RL in a data science project

Understand today’s limitations of Reinforcement Learning and the main challenges to be solved in the field

Content

Definition of Reinforcement Learning in the context of the wider field of Machine Learning: Specificities of RL compared to supervised and unsupervised machine learning

Examples of Reinforcement Learning applications, overview of the major actors in the field and of the open-source frameworks for Reinforcement Learning

Presentation of the Key concepts in RL: Agent, Environment, Reward, Policy, Value-Function and State-Value Function, Learning vs Planning, Prediction vs Control, Exploration vs Exploitation, On-policy and Off-policy

Introduction to the most-used Reinforcement Learning Algorithms: dynamic programming, Monte-Carlo Algorithm, TD Learning, Multi-Armed Bandit Problem, Deep-Q network

Implementation of simple RL problems in python using Open AI Gym such as the frozen lake problem, a black-jack game, taxi scheduling, etc

Additional resources to dive deeper into the exciting field of Reinforcement Learning