[mlpack-git] [mlpack] Bug fix for Elkan? (#592)

[Erich Schubert]

Moving the MLpack related discussion from https://github.com/zoq/benchmarks/issues/9 here:
Yes, the PR is specific to Elkan; I believe at least in the dual tree code it was only resetting the one affected centroid. However, all algorithms may benefit from by default leaving empty centroids in place instead of pushing them to `DBL_MAX`. Often, these clusters will not remain empty longer than one iteration.
Furthermore (also not in this pull request), initialization by choosing k random objects as centroids is both very easy to implement, and much more effective than the current approach (which yields near-identical centroids on large data).

---
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/mlpack/mlpack/pull/592#issuecomment-201273076
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack-git/attachments/20160325/c2eb4a46/attachment.html>