[mlpack-git] [mlpack] mlpack KMeans class much slower than armadillo kmeans() (#514)

[Ryan Curtin]

I kept thinking about this, and I realized this is an unfair comparison, because if you aren't starting from the exact same starting centroids, the algorithm may converge in different ways.  If you specify random starting points and use `arma::keep_existing`, that will be a fairer comparison.  I also somehow missed the "without openmp" numbers; sorry about that.  It's possible that all the abstractions used for mlpack's k-means algorithm cause some amount of slowdown.  It should be avoidable, but it may be a while until I get around to improving it (if you feel so inclined, I'm happy to help with the contribution!).

---
Reply to this email directly or view it on GitHub:
https://github.com/mlpack/mlpack/issues/514#issuecomment-178758146
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack-git/attachments/20160202/4db5b1af/attachment.html>