/FlappyBirdwithDDRL

Implement the DDRL(Dueling Deep Reinforcement Learning) algorithm in the game called Flappy Bird

Primary LanguagePython

Playing Flappy Bird Using Dueling Deep Q Net

I rewrite and modified the code from another repo which is built to implement the classical DQN algorithm, came out in Nature 2015, in Flappy Bird Game

To run the code, just type python FlappyBirdDQN.py

About the code

If you want to use the classical DQN architecture, just change the Dueling statement to be 'false' in line 24, FlappyBirdDDQN.py

The saved network is not published because the upload speed is limited. It'll take a long time to upload them since the size of the saved parameters of the network architecture is large.

If you are interested in the well-trained network please email me at: yangshuo960502 at foxmail dot com

Disclaimer

This work is based on the repo: (https://github.com/floodsung/DRL-FlappyBird)