[mlpack-git] [mlpack/mlpack] R+ and R++ trees implementation (#699)
Ryan Curtin
notifications at github.com
Wed Jun 29 16:36:55 EDT 2016
Looks good to me. I added some comments; when they are addressed I don't think I have any more comments. A couple of high-level questions:
* We had decided that the R+ tree was deficient and could not be implemented in a way that avoided overlap, but you've implemented it here anyway. It looks to me like you used a heuristic that will try to avoid overlap but can sometimes fail, and in that case you try to minimize overlap. Am I understanding that correctly? I think it's fine to include the R+ tree like this, but we should document that it is modified from the original paper implementation.
* What happens if I build a tree on all the same point? Like if I have a dataset of 100 points all at the origin. Does this still give a valid tree?
* We should notify the R++ tree authors that we have implemented their method when this is merged, and maybe provide a couple of quick performance comparisons.
I'll do a few performance comparisons as we get closer to merging this. If you want, we can add support for the R+ and R++ trees to NSModel in this PR. That's up to you if you want to do that now or later.
---
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/699#issuecomment-229480524
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack-git/attachments/20160629/ca08f872/attachment-0001.html>
More information about the mlpack-git
mailing list