MATLAB: Kmeans function give us give different answer

clusteringk-meanskmeansStatistics and Machine Learning Toolbox

I have noticed that kmeans function for one k value in a single run gives different cluster indices than while using in a loop with varying k say from 2:N. I do not understand this. It will be great if it is clear to me.

Best Answer

Because, if you are using the default settings, kmeans() randomly selects a starting point. The algorithm is not deterministic and the results might depend on that starting position.

Related Solutions

MATLAB: Kmeans explanation of information obtained

IDK is the indices that match each of the N rows of input into a cluster number.

C are the locations of the centroids of the k clusters, each of which will have size(B,2) components.

D returns the distance between each of the N input rows and each of the k cluster centers. So this would be pdist2(B, C)

sumd is a cluster statistic that I am unsure of the use of. Numerically it is

for K = 1 : k
  sumd(K) = sum(D(IDX == K, K));
end

that is, for each cluster, it is the sum of the distances between the points that belong to the cluster and the center of that cluster.

MATLAB: Help in k means clustering

There is a randomness component in k-means clustering, as you surmise.

Have you tried putting the statement

rng(1)

at the beginning of your code? This will set the generator seed, and therefore produce the same series of pseudorandom numbers each time you run it.

Best Answer

Related Solutions

MATLAB: Kmeans explanation of information obtained

MATLAB: Help in k means clustering

Related Question