0

Does the SOM clustering algorithm have cluster centers? If there is how to use matlab's SOMtoolbox package to calculate this cluster center.

I observed that the kmeans_clusters clustering function in the somtoolbox package can return the cluster centers. The code is as follows:

function [centers,clusters,errors,ind] = kmeans_clusters(sD, n_max, c_max, verbose)

where 'centers' are the cluster centers

I tried the following code, may I ask if the 'sM.codebook' here is the cluster center?

clusterNum = 99;
sD = som_data_struct(data);%首先可以判定能出图，所以这个不用转置
sM = som_make(sD, 'msize', [clusterNum],'training', 'long');
centroids = sM.codebook;
save('centroids.mat','centroids','-v7.3');
Bmus = som_bmus(sM,sD);
mr yu
  • 1
  • Note that the SOM Toolbox is not an official MATLAB toolox, but a [third-party software](http://www.cis.hut.fi/somtoolbox/) – Adriaan Mar 01 '22 at 12:31
  • I don't know the package itself, but in general, SOMs use "soft" clusters without centers (compared to k-means, which uses "hard" clusters). – felice Mar 01 '22 at 19:16

0 Answers0