[mlpack-git] [mlpack/mlpack] Vantage point tree (#708)

[Ryan Curtin]

I think you're right that `FirstPointIsCentroid()` will be no longer applicable.  The whole idea of that is to avoid extra calculations of the center-to-center distances between nodes.  If we have another child for only the vantage points, we can still avoid center-to-center distance calculations, but it's less clear precisely how to do that.  The extent of my thinking so far is that we can define some trait like `FirstSiblingFirstPointIsCentroid` or something like this, that indicates that `Parent()->Child(0).Point(0)` is the center of the node.  But it's less clear how we would, e.g., cache the center-to-center calculation for that sibling.  I guess it is possible to use `traversalInfo` for that, if we can assume that the first sibling is always recursed into before the other siblings.  (That last bit might be another tree trait we can add, although really that's a traversal trait, so maybe this implies a new traits class?  Or alternately we can just check in the `traversalInfo` object to see if that sibling was the previous base case.)

---
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/708#issuecomment-234604538
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack-git/attachments/20160722/1eea5383/attachment.html>