https://github.com/jalajthanaki/q_learning_for_simple_atari_game
This repository contains the code which can help us to understand how q-learning algorithm can be applied to build simple video game bot.
https://github.com/jalajthanaki/q_learning_for_simple_atari_game
atari qlearning-algorithm reinforcement-learning-algorithms
Last synced: 5 months ago
JSON representation
This repository contains the code which can help us to understand how q-learning algorithm can be applied to build simple video game bot.
- Host: GitHub
- URL: https://github.com/jalajthanaki/q_learning_for_simple_atari_game
- Owner: jalajthanaki
- Created: 2018-02-22T07:55:59.000Z (over 7 years ago)
- Default Branch: master
- Last Pushed: 2018-02-23T09:53:24.000Z (over 7 years ago)
- Last Synced: 2025-03-31T22:21:52.680Z (6 months ago)
- Topics: atari, qlearning-algorithm, reinforcement-learning-algorithms
- Language: Python
- Size: 5.86 KB
- Stars: 3
- Watchers: 1
- Forks: 2
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Q learning for simple atari game
This is a simple example of a type of [reinforcement learning](https://en.wikipedia.org/wiki/Reinforcement_learning)
called [Q learning](https://en.wikipedia.org/wiki/Q-learning).
## Overview
We are building simple game and using Q-learning algorithm we built the bot which can able to will this simple game
● Rules: The agent (yellow box) has to reach one of the goals to end the game
(green or red cell).
● Rewards: Each step gives a negative reward of -0.04.
The red cell gives a negative reward of -1.
The green one gives a positive reward of +1.
● States: Each cell is a state the agent can be.
● Actions: There are only 4 actions. Up, Down, Right, Left.
## Dependencies
- Python 2.7
- tkinter
## Installation
- To install tkinter You need to execute this command: ```$ sudo apt-get install python-tk```
## Usage
Run `python Learner.py` in terminal to see the the bot in action. It'll find the optimal strategy pretty fast (like in 15 seconds)
## Credits
The credits for this code go to [PhillipeMorere](https://github.com/PhilippeMorere) and [joongwha](https://github.com/joongwha). I've merely created a wrapper to get people started.