Snake Ai - using deep Q learning

This project is about using a branch of reinforcement learning (deep Q learning) to teach an AI how to play the game snake.

NOTE: The code for the game.py and model.py files was modified from the repo https://github.com/patrickloeber/snake-ai-pytorch/.

Here is a blog post explaining the underlying algorithm (unfinished...)

Set up

1.  git clone https://github.com/VishalJ99/snakeAI.git	
   

2.  cd snakeAI
   

3.  python -m venv snakeAI    	# mac virtualenv snakeAI
   

4.  snakeAI\Scripts\activate.bat # mac source snakeAI/bin/activate 
   

5.  pip3 install -r requirements.txt
   

6.  python agent.py
   

It will take about 150 episodes / 10 minutes of training before the agent starts reliably getting scores > 10.