Is the correlation distance function for training self-organizing map (SOM) available in Matlab?

Hi.

I have a data set of expression levels of different genes, and I want to build a SOM to cluster different gene expression levels. Now I really need to use correlation distance metric to train the map.

I noticed that in Matlab there are four distance functions are available : dist, linkdist, mandist, boxdis. Normally, dist (refers to Euclidean distance) is widely used. And also, I can find the correlation distance can be obtained by using D = pdist(X,'correlation') .

Is it possible that train a SOM using correlation distance function in Matlab?

Thanks.

[eng keong]

Hi there, would like to know how do you update your distant when you use Correlation?

Answer 1

[Amith Kamath]

I don't know if MATLAB has in-built functions to handle those aspects of SOM, but this may help:

http://www.cis.hut.fi/somtoolbox/package/docs2/somtoolbox.html

Thanks!