Detailed description |
Transferred from EPCC TRAC tracker-
Submitted on 27 Apr 2012
"Did you consider implementing the pamonce option (see http://stat.ethz.ch/R-manual/R-devel/library/cluster/html/pam.html), that has been added recently to pam. In my experiments, the value of “true” gave a runtime benefit of roughly 30% with identical results- I did not try pamonce=2 yet."
Michal's thoughts on this were
"I guess there is no option to do this at the moment, but I'm not sure how to respond to Georg.
The pamonce option implements an algorithmic short cut into pam, with optimised swap operation. The process of calculating the cost of removing or adding new medoids is more local I think.
I've looked through the modifications to the original pam code, this was updated Jan 2012, and it looks like the new algorithm would be fairly straightforward to port to SPRINT. It should definitely improve the performance, but I can't guarantee the 30% as this part is already heavily optimised in ppam." |
|