AlphaBulletScreen

Play Bullet Screen Game with Artificial Intelligence

Features

• Complete Bullet Screen Game based on pure Python

• Deep reinforcement learning [1]:

  Home-made Deep Q network

How to use

• Play BulletScreen Game by yourself

    cd alphabulletscreen
    python alphabulletscreen.py --play --verbose
  

• Re-train AI model for BulletScreen Game

    cd alphabulletscreen
    python alphabulletscreen.py --retrain --verbose
  

• Continue to train AI model for BulletScreen Game

    cd alphabulletscreen
    python alphabulletscreen.py --train --verbose
  

• Watch the AI model to play BulletScreen Game (Please train AI model first)

    cd alphabulletscreen
    python alphabulletscreen.py --playbyai --verbose
  

The option "--verbose" is for debugging purpose
Press key left,right,up and down to play the game

E-mail

[email protected]
You're most welcome to contact with me to discuss any detail about this project

Extra

I have written many home-made mini-game projects for mastering the reinforcement learning technologies, which can be found in my GitHub repositories. Maybe there are some hidden bugs behind them, i.e, game mechanism flaws, ill network structure, unreasonable or unappropriate parameters. But I am making progress and try to do my best. Practices make better. -- Yang Long (2018.9.20 GMT+8)

Reference

[1]Mnih, Volodymyr, et al. "Human-level control through deep reinforcement learning." Nature 518.7540 (2015): 529.