Epistasis Blog

From the Computational Genetics Laboratory at the University of Pennsylvania (www.epistasis.org)

Monday, May 26, 2008

MDR 1.2 Released

We have released version 1.2 of our open-source Multifactor Dimensionality Reduction (MDR) software package. It can be downloaded from here.

This new version includes interaction graphs to complement the existing interaction dendrogram for visualizing interaction information as a way to statistically interpret MDR models and attribute relationships. See our 2006 paper in the Journal of Theoretical Biology for more information about interactions graphs and dendrograms. See also my more recent review in the book Knowledge Discovery and Data Mining (email me for a pdf version). This work is inspired by the excellent work of Dr. Aleks Jakulin. See his list of papers on the topic here.

For more information about MDR please visit www.epistasis.org and www.multifactordimensionalityreduction.org. Also see the November and December 2006 posts on this blog for the MDR 101 tutorial. Let us know if you have any questions, comments or concerns.

We are planning to release MDR 2.0 this summer that will include estimation of distribution algorithms (EDAs) for stochastic searching.

2 Comments:

At 2:35 PM, Blogger nojhan said...

Working on EDA (and other metaheuristics) myself, its always interesting to see other researcher use them. I'm curious to know how you plan to use these algorithms...

* Firstly, how did you heard about EDA?
* Why choose stochastic optimization algorithms instead of deterministic and/or complete ones?
* Do you plan to use a specialized framework or to reimplement your own methods?
* And finally, how did you plan to validate your final approach?

Nice software, by the way.

 
At 11:30 AM, Blogger Jason Moore said...

* Firstly, how did you heard about EDA?

I have worked in the evolutionary computing area for some time and was exposed to EDA through that community,

* Why choose stochastic optimization algorithms instead of deterministic and/or complete ones?

The search space is highly nonlinear and astronomical in size.

* Do you plan to use a specialized framework or to reimplement your own methods?

Our own methods that are based on ant colony optimization (ACO) which is a simple EDA.

* And finally, how did you plan to validate your final approach?

We just published a paper at the ANTS'08 conference on using ACO that shows it is an effective strategy in this domain. We rely heavily on simulation studies to validate these approaches.

* Nice software, by the way.

Thanks! Feedback welcome.

 

Post a Comment

<< Home