[R] Hierarchical Clustering Using Mutual Information

[Martin Maechler]

>>>>> "Julio" == Julio Thomas <julio_semprones at yahoo.co.uk>
>>>>>     on Fri, 9 Dec 2005 21:13:20 +0000 (GMT) writes:

    Julio> Dear R-helpers, Is there somebody who knows if R has
    Julio> already a build in function for Hierarchical
    Julio> Clustering which uses Mutual Information as proximity
    Julio> measure?
   
The most prominent hiearchical clustering functions in R,
hclust() and agnes() and diana() {package 'cluster'}, are based
on general dissimilarity "matrices";

so you just need to be able compute  
   d(i,j) <- Mutual_Information( unit[i], unit[j] )
and then use for example hclust().

If 'MI' was a matrix with these numbers, you'd use

    dMI <- as.dist(MI)
    hc.res <- hclust(dMI, ....)
    
    plot(hc.res, .....)
    ....

Martin Maechler, ETH Zurich