Error in kmeans of Matlab R2013 ?

Hello

I have an unexpected error by running kmeans in the last version of matlab 2013. I tried to cluster a data matrix having 12410 rows and 15 columns (it is actually Principal Component). There are no missing entries in the dataset. When I type >> k=kmeans(PC,2);

I have the following error

Error using randsample (line 106) REPLACE must be either true or false.

Error in kmeans/loopBody (line 357) C = X(randsample(S,n,k),:);

Error in internal.stats.parallel.smartForReduce (line 128) reduce = loopbody(iter, S);

Error in kmeans (line 295) ClusterBest = internal.stats.parallel.smartForReduce(...

I precise that I had not this error on previous version with exactly the same coding

Thanks in advance for any help and suggestion

Vincent

Answer 1

[Ilya]

Type

which kmeans
which randsample

to verify that you are using the functions from the official MATLAB installation. If you are, try to reproduce the problem on a smaller set of data and post that set here.

[Vincent Moron]

You were right. I had an old version of randsample stored in my personal directory and it mixes with the new one.

Thanks for the quick reply

Vincent

[Veera]

My matlab is using latest kmeans and randsample. There is no problem on small dataset and my data is output of PCA. I am facing this issue on larger dataset. Size of the dataset is around 250,000 vectors (250,000 * 188).