Deep Reinforcement Learning for Flappy Bird
See our video demo on YouTube Video Demo
Want to train a Flappy Bird?
Deep Q Learning algorithm is originally described in Playing Atari with Deep Reinforcement Learning, a paper from NIPS 2013 Deep Learning Workshop from DeepMind.
In this repository, instead of Atari games, we try to play with Flappy Bird. We use an open source JavaScript library ConvNetJS to train Deep Learning Models.
Our implementation follows Deep Q Learning Demo
Our goal is to train a deep neural network with q-learning technique to learn control policy from inputs generated by the game. Over time, the flappy bird learns to flap or not at a point to avoid as many pipes as possible.