https://github.com/grypesc/artificialsnake

Snake game in python with intelligent, self learning agents.
https://github.com/grypesc/artificialsnake

agent artificial-intelligence artificial-neural-networks intelligent-agent intelligent-agents neural-network snake snake-game snakegame

Last synced: 18 days ago
JSON representation

Snake game in python with intelligent, self learning agents.

• Host: GitHub
• URL: https://github.com/grypesc/artificialsnake
• Owner: grypesc
• License: gpl-3.0
• Created: 2019-05-24T06:49:12.000Z (over 5 years ago)
• Default Branch: master
• Last Pushed: 2019-07-15T18:10:30.000Z (over 5 years ago)
• Last Synced: 2024-10-28T17:12:52.255Z (2 months ago)
• Topics: agent, artificial-intelligence, artificial-neural-networks, intelligent-agent, intelligent-agents, neural-network, snake, snake-game, snakegame
• Language: Python
• Homepage:
• Size: 16.1 MB
• Stars: 4
• Watchers: 1
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# artificialSnake

Snake game in python with different types of agents.

  

![](gifs/selfLearningAgent.gif)  

Self learning agent learns on its own. Each step is added to a dynamic dataset and the move is predicted by a neural network. In the next step the previous step is evaluated by an evaluation function. Every generation learns only from those datapoints that were well evaluated (evaluation was higher than a threshold). Once per 4 generations 10% of the oldest datapoints is deleted. What's more if the neural network predicton has a low prediction probablility than the snake makes a random moves. Just like a human it can behave randomly and then learn from mistakes.

  

![](gifs/datasetAgent.gif)  

Dataset agent was trained on a dataset created by me, it mimics my moves but cannot overcame entering places where there is not enough space and doom is brought to him in few next moves.