You can do it!!! Top 5 resources to EASILY learn Deep Reinforcement Learning

Summary

The undefined website provides a curated list of top 5 resources for learning Deep Reinforcement Learning (DRL), emphasizing the ease of finding quality online materials compared to other research areas.

Abstract

The author of the article on the undefined website shares their personal journey from DRL novice to researcher, leveraging the abundance of online resources. They recommend a structured approach to learning DRL, starting with the Hugging Face DRL course, which combines practical coding with theoretical understanding, and progressing through a series of books and papers that cover both practical tools and theoretical foundations. The resources include a mix of interactive courses, books that delve into the theory and practice of DRL, and a seminal survey paper that provides an overview of the field. The article emphasizes the importance of understanding both the implementation and theoretical aspects of DRL to effectively contribute to the research area.

Opinions

The Hugging Face Deep Reinforcement Learning course is highly praised for its comprehensive content, interactive leaderboards, and the use of Jupyter Notebooks that can be run on Google Colab.
"Foundations of Deep Reinforcement Learning: Theory and Practice in Python" is recommended for its readability and introduction to SLM Lab, another tool for training DRL bots.
OpenAI's "Spinning Up in Deep RL" is valued for its guidance on entering the DRL research area and its list of research topics and benchmarks.
The Reinforcement Learning book by Sutton and Barto, although more theoretical and less focused on deep learning, is considered essential reading for understanding the foundations of the field.
The survey paper by Li, Y. (2017) is suggested for gaining a broad perspective on current research trends in DRL and for helping to select a specific research direction.
The author expresses enthusiasm for the fun and engaging nature of DRL research and encourages reader interaction through comments and subscription to similar posts.

You can do it!!! Top 5 resources to EASILY learn Deep Reinforcement Learning

One year ago, I knew almost nothing about this exciting area of research when I became aware of its awesome potential. Now, I am involved into 3 lines of research dealing with Deep Reinforcement Learning. How did I do this? I briefly explain in this medium article.

Learning Deep Reinforcement Learning is nowadays easy because we have a lot of available resources online. Here is a brief introduction to them.

When I began my PhD on Bayesian optimization I had to learn about the topic reading related books and papers, what was really a hard process. Concretely, Bayesian optimization was a hard topic, and few resources talking about it were available online. However, it is exactly the opposite with Deep Reinforcement Learning. In particular, a lot of resources are available online. Ironically, it is necessary to provide not an exhaustive but a curated list to not waste the time in useless resources.

Top 5 Resources

I suggest you to learn in the following order, that also represents how good I consider the resource to learn about this exciting topic.

Hugging Face Deep Reinforcement Learning course: Possibly the, believe me, best course I have ever taken. Amazing work by Thomas Simonini, you will learn simultaneously how to code and use DRL algorithms using StableBaselines3 and OpenAI Gymnasium environments and the theory behind Deep Reinforcement Learning. And it has been so much fun! The course includes leaderboards where you upload your agents and they compete in the environments with the agents of other players. Almost all the code are Jupyter Notebooks that can be uploaded to Google Colab. It is insanely challenging to enter the top 10 in the different environments. They also have a Discord channel where you can chat with the other students to discover how they have trained the policy of the agent. I honestly think that we do all have to learn about this course in order to make teaching fun, awesome content!

2. Foundations of Deep Reinforcement Learning: Theory and Practice in Python: I have included this book in this position because it has a critical advantage, it is very easy to read and you will learn a different tool, SLM Lab, that you can also use to train bots. The idea is to first complete the previous course and, then, you can start with this book. If you do this, then, you will be familiarized with all the current topics of Deep Reinforcement Learning.

Very alternative book of Deep Reinforcement Learning that I have personally liked. It is also useful to learn another DRL tool such as SLM Lab.

3. Welcome to Spinning Up in Deep RL by OpenAI!. Now that you have get the main insights of Deep Reinforcement Learning and know several tools where you can implement new algorithms, it is the turn to learn how to enter this amazing research area. And this is introduced perfectly in this webpage by OpenAI. And with a lot of research topics available, such as key papers in DRL and useful benchmarks. Amazing resource!

I honestly wish that all research areas would have something like this, it will dramatically boost research.

4. Reinforcement Learning book by Sutton and Barto. Not exactly DRL but, in this point, you will need some connections with other areas (and this book does that perfectly well with exciting chapters on psychology and neuroscience) and more formalism. I recommend you this book for you to not forget the foundations behind reinforcement learning, but it is an advanced, less practical but more theoretical book. However, it is a classic, and because of that, I think that you must read this if you are going to work in this area.

A classic. It must be included in this top 5, although it is not **Deep Reinforcement Learning** exactly, but must be read. Exciting connections with other areas.

5. Li, Y. (2017). Deep reinforcement learning: An overview. arXiv preprint arXiv:1701.07274. Finally, you need to choose your research line, and need a broad view about what is being done in deep reinforcement learning. This survey paper is great at that. I recommend you to read it and then search for more recent research on your particular direction.

Lots of luck with the process!!! I am enjoying a lot this line of research mostly because it is very fun! Please leave me in the comments what do you think about this top and if you like these posts, subscribe!!!