Hierarchical clustering with similarity matrix

Hi there,

I have a similarity matrix that I would like to use as the input of the function linkage. However, this takes as an input only the a dissimilarity/distance matrix. Do you know any way by which I can manage to run a hierarchical clustering in Matlab using my similarity matrix?

Cheers!

Answer 1

[Akanksha Shrimal]

Hi,

It is my understanding that you want to compute Hierarchical Clustering with ‘linkage’ function in MATLAB using a similarity matrix.

The ‘linkage’ function in MATLAB takes input as distance vector to compute clusters.

Given a similarity matrix s, we can compute the dissimilarity matrix of s by

ds = 1 – s

This dissimilarity matrix is then converted into a vector form that 'linkage' accepts using 'squareform'.

X = squareform(ds)  

Now you can apply 'linkage' with X using ‘complete’ method for calculating distance between clusters.

Z = linkage(X,’complete’) 

For more details refer to this documentation link.

Hope this helps.

[Umar]

Hi @Akanksha Shrimal,

I executed the code above following your mentioned instructions,”Given a similarity matrix s, we can compute the dissimilarity matrix of s by

ds = 1 – s

This dissimilarity matrix is then converted into a vector form that 'linkage' accepts using 'squareform'.

X = squareform(ds) “

and encountered an error in code. Please see attached.

Am I missing something?

The definition of similarity matrix can be found here,

https://www.mathworks.com/help/stats/spectralcluster.html#mw_319da951-f871-482c-b648-5056f4ebff46_sep_shared-SimilarityMatrix