using kmeans function in parallel

Hi, I want to use kmeans function in parallel, with a matlabpool of workers. When I open a matlabpool and then execute kmeans with Useparallel = true, only 1 core is working. The default numbers of workers is 4. What is wrong? Should I put the execution of the kmeans function inside a spmd block? Thanks in advance,

Jordi

[Marwan Radwan]

please i need this code and steps how to set useparallel i am new in matlab thank you alot

Answer 1

[Shashank Prasanna]

Hi Jordi, when the 'useparallel' flag set to true, KMEANS only computes the replicates in parallel.

If you have your replicates set to 1 you are unlikely to see a performance difference.

Try setting a larger 'replicates' to compare the performance difference.

Replicates are used to run kmeans several times with different start point so as to have a better chance of finding a global minimum and better centroids.

Answer 2

[Thomas]

Don't know if this would help but you can find an implementation of k-means in MATLAB using MATLABmpi (message passing interface) http://www.cs.toronto.edu/~dross/code/parallel.shtml

[Abeera Tariq]

Hi, do u have knowledge of matlabmpi?

Answer 3

[Jordi]

Thanks a lot !

Jordi