Reinforcement Learning - Learning how to play Flappy Bird in PyTorch

This sample is published as part of the corresponding blog article at https://www.toptal.com/deep-learning/pytorch-reinforcement-learning-tutorial

Visit https://www.toptal.com/blog and subscribe to our newsletter to read great posts!

Dependencies:

• Python 3.6.5
• Run pip3 install -r requirements.txt to install dependencies.

How to run:

• Run python dqn.py test to run pretrained neural network model.
• Run python dqn.py train to train the model from the beginning. You can also increase FPS in game/flappy_bird.py script for faster training.

References:

• https://awesomeopensource.com/project/sourabhv/FlapPyBird
• https://awesomeopensource.com/project/yenchenlin/DeepLearningFlappyBird -> modified FlapPyBird game engine adjusted for reinforcement learning is used from this TensorFlow project
• https://ai.intel.com/demystifying-deep-reinforcement-learning/
• https://www.cs.toronto.edu/~vmnih/docs/dqn.pdf
• https://storage.googleapis.com/deepmind-media/dqn/DQNNaturePaper.pdf
• http://cs229.stanford.edu/proj2015/362_report.pdf
• https://en.wikipedia.org/wiki/Convolutional_neural_network
• https://en.wikipedia.org/wiki/Reinforcement_learning
• https://en.wikipedia.org/wiki/Markov_decision_process
• http://www0.cs.ucl.ac.uk/staff/d.silver/web/Teaching.html
• https://en.wikipedia.org/wiki/Flappy_Bird
• https://pytorch.org/