[mlpack-git] [mlpack] Bugfix #350 and some fixes in RStarTreeSplit and XTreeSplit (#556)

[lozhnikov]

I looked through the X-tree paper. The authors don't explain the choosing subtree algorithm. But according to the R*-tree paper, alternative methods didn't outperform Guttman's original algorithm (for choosing the best non-leaf node) and for the leaf nodes, minimizing the overlap is better.

However,  the main idea of X-tree is to reduce overlap in the split algorithm. Why do not use the same idea? For example, 
1. Try the R* algorithm.
2. If the overlap cost > MAX_OVERLAP, try to choose  the node with the minimum enlargement and overlap cost < MAX_OVERLAP.

But it is my assumption, I haven't any benchmarks yet.

---
Reply to this email directly or view it on GitHub:
https://github.com/mlpack/mlpack/pull/556#issuecomment-194480441
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack-git/attachments/20160309/5deb70c2/attachment-0001.html>