
marioai
Experiments in applying evolutionary algorithms, neural networks and other AI/CI/ML algorithms to Super Mario Bros.
MarioAI is a benchmark for machine learning and artificial intelligence based on Super Mario Bros.
Check out the running Mario AI Championship 2010 at http://www.marioai.org !
NEW: Turing test track on CIG 2012 Coming!
What?
This is an open source software for AI, Machine Learning algorithms benchmarking, written in Java. Coupled with AmiCo library allows agents and learning algorithms implemented in Java, C++, Python. This software suits for research, teaching and self-learning. Starting from very simple examples you have ability to dive into very complex tasks. JNI is going to be supported fairly soon.
Why?
Got bored another day from walking just the mazes or balancing even the double pole. Came up with idea to create a universal benchmark with simple interface and entertaining content.
When?
This project was started in March 2009. Mario AI Competition 2009 took place @ ICE-GIC in London on August, 24th and CIG in Milano on September, 7th, 2009.
Currently approaching and working on organizing AI Mario Championship 2010 at
EvoStar 2010, Istanbul WCCI 2010, Barcelona CIG 2010, Copenhagen GIC 2010, Hong Kong
Who?
by Julian Togelius and Sergey Karakovskiy with participant of Tom Schaul and Jan Koutnik.
Where?
@ IDSIA, Switzerland
@ SPBSU http://apmath.spbu.ru, Saint-Petersburg State University, Russia
@ TSURE, Taganrog State University, Russia
@ ITU, IT University of Copenhagen, Denmark
Project Information
- License: New BSD License
- 41 stars
- svn-based source control
Labels:
AI
artificialintelligence
machinelearning
benchmark
SuperMarioBros
Java
Python
automaticlevelgeneration
marioaicompetition
marioaichampionship
AmiCo
interlanguagecommunication
javapythoncommunication
MarioAIChampionship
scalablereinforcementlearningbenchmark