These are my notes taken in the Fundamentals of Reinforcement Learning course by the University of Alberta in Coursera, part of their Reinforcement Learning Specialization. It is a very interesting framework of learning not only because it is very game-like, but because it helps modeling the world when you want to be able learn from sequences of interactions which progressively expand the model of the world and the actors in it, differing a lot from traditional supervised and unsupervised learning.

This course (and the whole specialization) teaches a lot of theory following the famous Sutton & Barto book, but they also offer a platform in which you can validate your learnings by implementing the algorithms and seeing how they perform.

I’ve been working as a Software Developer supporting Data Scientists in their creation of Machine Learning models, and most recently specificly Reinforcement Learning models, they do the modelling and I work on the backend integration of such models to use them in production, creating data pipelines and maintaining their cleanliness, etc. Still, I wanted to learn more about what they were doing so that I’d be able to provide the best support from my side and even think of ideas to implement.