Now that we have a working installation of ML-Agents, we can take a more hands-on look at the system.
In this chapter, I’ll discuss the various components of ML-Agents. We’ll then discuss the learning environments inside the 3DBall sample scene, in which ML-Agents aim to balance a ball on their heads.
From here, we’ll talk about inputs and outputs and get a better look at how our agent views the simulated world. Then we’ll talk about rewards more practically and in depth than earlier in the book.
We’ll ...