The Daily Click

::. ::.

Project: DringleTopia Artificial Intelligence Collection

Project Started:	29th June, 2009	Last Update:	12th July, 2009
Project Owner:	CodeCannon	Project Members:
Project Type:	Extension	Project Progress:
More Info:	http://www.dringletopia.da.ru/projects/dringletopia-artificial-intelligence-collection	Faves:	4

Project Overview
The DringleTopia Artificial Intelligence Collection (DAIC) is a small group of extensions for MMF2 that allow you to implement neural networks, automatic problem solvers and a few other types of artificial intelligence into your games and utilities. Clickteam forum thread: http://www.clickteam.com/epicenter/ubbthreads.php?ubb=showflat&Number=150095#Post150095

Project Forums	Topics	Posts	Last Post
Public Forum Public forum for open project discussion	1	6	AI Experiments By: CodeCannon On: 12th Jul 09, 22:07:06
Private Forum Private forum for project members only	0	0	No Access

In the works

Posted 12th Jul 09, by CodeCannon

Post A Comment

I've been writing up an introductory tutorial that shows you how to make a simple ping pong game with an AI opponent. Likely it'll be done by the end of July.

Also, I've been perusing Prof. Jürgen Schmidhuber's interesting page ( http://www.idsia.ch/~juergen/ ), and I like the idea of OOPS ( http://www.idsia.ch/~juergen/oops.html ). Hopefully it can be used in MMF. Just think - in a small way you might be able to get MMF to program itself!

First public release (0.2.1 A)

Posted 29th Jun 09, by CodeCannon

1 Comment

The first of the collection is called Static ANNs w/ Genetic Algoritm - DAIC. It is currently in its alpha phase.

Notable features
- Real-time input/output and score history graphs
- MMF2 debugger support
- Many run-time configurable settings
- Each object can handle a population of agents, not just one

Wish list / Task list
- Distributed training / training farm support (think render farm) using OINC / MOO Click
- Error handling
- Rote escape (helps genetic algorithm get away from a local minima)
- Implement settings' actions/expressions

Technical Specifications
- Multilayer perceptron
- Long short-term memory
- "Enforced sub-populations" genetic algorithm
- Topologically static (fixed network structure)
- "Double" data type

Enjoy!

View Project Archive

Project Forums

Public Forum

Quick Links

Project Main
Project Archive
Project Forum
Project Members

Favourite

Is this worth a look? Let others know!

Favourite

Worth A Click