Spinning Up a Pong AI with Deep Reinforcement Learning

This seems like a solid example of practicing Deep RL with a few lines of code. With the OpenAI thing you run a program from command-line. Curious if you have any insight about what the algorithm is and how it compares to the Keras one?

Both the Keras one and the one from spinning up can be called "vanilla policy gradients". The one in Keras is closer to REINFORCE, and the one from spinning up use actor-critic and a multilayer perceptron.

I’d like to see this AI hooked up to an actual Atari machine somehow. Has anyone tried something like that? Could the model process the frames quickly enough to move a robotic arm up or down on a joystick?

Hum, have not tried that but I guess the robotic part and the computer vision would be pretty standard, so it would be feasible.

PS:They even have robots playing real Pong! https://www.youtube.com/watch?v=tIIJME8-au8

Great post! You mentioned different RL frameworks in the article (Dopamine, Spinningup, Keras-RL, etc...) Which one do you recommend in term of usability?

Thanks! I would totally recommend Spinningup, because of the guide and the docs.