I implemented LSTMs and reinforcement Learning to train an AI who learns to play Breakout .
I was highly inspired when Google successfully made an AI which plays the Breakout using A3C . The Asynchronous Actor Critic and LSTM combines together to create an AI which doesn't take weeks to train .
This was a very difficult project for me as I am not an expert neither I can read research papers effortlessly . But after hours of watching tutorials , reading blogs , reading research papers and sleepless nights , I completed it .
The model.py contains all the ConvNet layers , Actor-Critic layers and LSTM layers .