Teaching Neural Nets to Fight with RL [P]

In this project I wanted to see if we could discover any interesting emergent behaviors if we trained two agents to play a streetfighter-like game.

It turns out that the agents are really good at reward hacking, but using a little reward shaping and league play I was eventually able to get something relatively interesting.

You can read the article and try it out for yourself here: https://blog.lukesalamone.com/posts/fighting-game-rl