Hi Wei Tai,
thanks, with your hint to rjava library, I did further investigate and found a solution.
Seems there was an issue with R 64-bit vs. java 32-bit and java 64-bit. I installed the latest version of java 64-bit, install.packaged() all the libraries in the list again (most of them were there, some seemed to be updated from 3.1.2 to 3.1.3.). Since paths in Windows and PA were already pointing to the correct folders, I did not have to change any more - R-k-means clustering was running in PA then :-).
I just wonder it was quite slow - e.g. 5 Clusters for 250 datapoints (only 2 continuous dimensions for each point) took like 1min in PA whereas in R you have like less than a second. Is there a lot of overhead when using R-functionality in PA?
Best Regards
Florian
