Step by Step Tutorial for Deep Reinforcement Learning Policy Gradients Method with Keras and OpenAi gym.
In this short project we are gonna train a neural network to play Pong game using a reinforcement learning algorithm (Policy Gradients Method - REINFORCE).
if you want to run it, just clone the repo and open the reinforcement_learning_pong_keras_policy_gradients.ipynb and read and run the notebook
we train a simple 200 hidden neuron network and a convolutional model.
sample playing of simple network:
video of simple network playing a game (https://www.youtube.com/watch?v=Ol163jSlEPI):
video of convolutional network playing a game (https://www.youtube.com/watch?v=1goeHG_hsUo):
Considering limited time and for learning purposes I am not aiming for a perfect trained agent, but i hope this project could help people get familiar with basic process of rl algorithms and keras. The above video took 3 days for agent to learn on a slow computer. to obtain production results, a lot of more training and tuning is required which is not our focus.
prerequisites: familiarity with neural networks,supervised learning, tensorflow and keras, openai gym