[mlpack] Implementing new trees to mlpack

Ryan Curtin gth671b at mail.gatech.edu
Mon Dec 22 11:39:23 EST 2014


On Mon, Dec 22, 2014 at 08:09:35AM -0800, Siddharth Agrawal wrote:
> Hello all,
> Myself Siddharth,
> I would like to introduce myself as I am new to mlpack organisation.
> I would like to start my work in implementing different tree types to
> mlpack library as presently there are very few trees implemented in
> mlpack.I am very much familiar with data structures & memory management. So
> I would like to get permission from mlpack before starting any kind of
> work. I would be grateful for this.
> Any kind of suggestion is deeply appreciated.

Hi Siddharth,

At first I thought you were the other Siddharth Agrawal but then it
became clear that we now have two. :)

Feel free to go ahead and start implementing trees.  When they are done,
you can submit them as a pull request on Github or as a patch on Trac or
as an email to this list (we are currently in a state of transition, so
for now, anything works...).

One tree that might be useful to start with is an octree which is
generalized to handle arbitrary dimensions.  It would probably be best
to make the dimension a template parameter (this should lead to a faster
implementation), and then make a wrapper around it so it can be used
with arbitrary datasets.

To get an idea of previous trees implemented for mlpack, you might look
at the BinarySpaceTree, CoverTree, and RectangleTree classes and imitate
their APIs.  If you need help, I am happy to help out. :)

Thanks,

Ryan

-- 
Ryan Curtin    | "How long have you had these things on?"
ryan at ratml.org | "Sixty-two years."


More information about the mlpack mailing list