#This submission is deprecated, please ignore it.
This is a simple convolutional network implementation that fits all atari games within the OpenAI Gym. It was implemented using AgentNet library in ~3 hours and trained on my colleague's GPU over ~4 hours All the details can be found
- SpaceInvaders here (stopped after 5, still not converged)
- AirRaid (stopped after 1k) in this notebook
Greetings!