[SciPy-user] Finding Neighboors

[Matthieu Brucher]

2007/8/1, Lou Pecora <lou_boog2000 at yahoo.com>:
>
> If you are using a data set that is static, i.e. you
> are not adding or deleting data points, then you
> should check out nearest neighbor searches using
> slices along various directions.  It's *much* easier
> to code than k-d trees and is about as efficient.  The
> reference is,
>
> Nene and Nayer, IEEE TRANSACTIONS ON PATTERN ANALYSIS
> AND MACHINE INTELLIGENCE, VOL. 19, NO. 9, SEPTEMBER
> 1997, pg. 989
>
> If your data is changing a lot as you are doing
> searches, then as I understand it, k-d trees are
> better to use since you can delete and insert nodes
> easily.


Isn't that just what kd-trees do ? If you take several slices around a
point, you must know which points belong to each slice, thus making a tree
anyway, no ?

Matthieu
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20070801/53df1f7b/attachment.html>