While browsing on the Spinning Up Open AI website, I was intrigued by the concept of Distributional Reinforcement Learning. What is it and what does it bring to the RL galaxy? Here, I expose the issues I faced when I implemented it and applied quantile regression DQN to Atari games.
In a previous post, I showed a DQN agent learned on a Unity environment to collect bananas. I used the Unity Machine Learning Agents Toolkit that allows us to build this kind of environment for Python. In this post, we will see how to build a learning environment and to use it.