Pong Agent using Policy Gradient

Contributors : Anhad Mohananey , Yashovardhan Chaturvedi

Description

The code in this repository will let you train a Convolutional Neural Network(CNN) to use OpenAI gym to play Pong solely using input frames of the game. The CNN is written in Keras. The code in pong_keras.py is based on Andrej Karpathy's blog on Deep Reinforcement Learning. You can play around with other such Atari games at the Openai Gym.

Setup

1. Follow the instructions for installing Openai Gym here. You may need to install cmake first.
2. Run pip install -e '.[atari]'.
3. Run python pong_keras.py