[mlpack] CMakeList adjustments for compiling Mlpack with Armadillo using openblas
Ryan Curtin
gth671b at mail.gatech.edu
Fri Jan 17 15:22:21 EST 2014
On Sat, Dec 28, 2013 at 08:30:34PM +0000, Steenwijk, Martijn wrote:
> I'll look into allkrann more closely, thanks.
>
> Considering ANN, I looked back into my code and ANNs manual, and it
> seems that it supports both bd trees as well as kd trees (I'm using kd
> trees). However, the splitting rule may be different compared to other
> implementations?
Sorry, I'm not sure why I didn't respond to this more quickly. The
mlpack splitting rule is mean-split; I'm not sure about the ANN
splitting rule. Sometimes median split is used to produce balanced
kd-trees (or at least, more balanced kd-trees). Mean split trees can
have weird degenerate cases where the right subtree has depth 1 but the
left subtree is very deep (or vice versa).
Thanks to Marcus's work, we'll have some benchmarks comparing FLANN and
ANN against mlpack for nearest neighbor search in a week or two (it
takes a very long time to run the mlpack benchmarks because they test
all the methods in numerous configurations). When they are updated I'll
respond to this post with a link and more info.
--
Ryan Curtin | "Maybe the next time."
ryan at ratml.org | - J.G. Ballard
More information about the mlpack
mailing list