[mlpack-git] [mlpack/mlpack] Major changes to solve issue: #275. (#624)

Naman Taneja notifications at github.com
Sun Apr 17 20:27:36 EDT 2016


Hi,
After a lot of brainstorming I found the problem in the implementation, which does not correctly handles the case when some nearSet elements are left (the while loop in create children: while(nearSet > 0)). I have thought of a solution to correct it.
It is based on the fact (which I would request you to confirm as well) that whenever a child node is created and childNearSet size and childFarSetSize are sent, the childNearSet will always be used (again the while condition: while(nearSet > 0)). The remaining nodes (which obviously belong to childFarSetSize) can redistributed to our near and far sets, so that we can continue properly executing the creation of tree.
To implement the changes we will need to update the SplitNearFarSet and PruneFarSet function, and create a new function that will put the remaining nodes in proper position.
Once correctly implemented the implementation of cover tree won't require and extra memory for childIndices and childDistances and any unnecessary memory movement will be reduced.

---
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/624#issuecomment-211138406
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack-git/attachments/20160417/81946cd6/attachment-0001.html>


More information about the mlpack-git mailing list