[mlpack-git] [mlpack/mlpack] Random projection trees (#726)
Ryan Curtin
notifications at github.com
Fri Aug 5 10:59:43 EDT 2016
Ok, I think that this is a good start. I like that you were able to reuse `BinarySpaceTree`. In addition to the comments I made inline, I think there are three main things left to be done:
* The `HRectBound` will not be a tight bound. Instead the true bound is a random polytope. But I am not sure it is feasible right now to figure out how to precisely bound that polytope, so we can open an issue and leave that for later.
* I need to run some timing tests to ensure that performance is reasonable with the RP tree. I think that it should be the same order of magnitude as the kd-tree search time for `mlpack_knn`.
* Can you think of any more tests? Right now the tests just ensure that the dataset is ordered reasonably. But I think it makes sense to test the `RPTreeMaxSplit` and `RPTreeMeanSplit` classes individually, and then maybe run a test where we ensure that there does exist some splitting hyperplane between each child. (I am not sure how easy that test would be to construct.)
---
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/726#issuecomment-237873628
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack-git/attachments/20160805/4306b6b8/attachment.html>
More information about the mlpack-git
mailing list