[mlpack] [GSOC 2016] Contribute for MLPack

Ryan Curtin ryan at ratml.org
Mon Mar 21 10:59:00 EDT 2016


On Sat, Mar 19, 2016 at 06:07:04AM +0530, Parijat Dewangan wrote:
> Hello,
> 
> 
>    - I have uploaded my draft for the project '*Implement tree types*' on
>    GSOC 2016 site. Please review it.

We will review it between the submission deadline and when decisions are
made.  It looks to me like your proposal is clear, so I have no further
advice to give.

>    - I have few doubts regarding the implementation of trees. Currently, I
>    have mentioned 5 trees which would be implemented in the project
> time - *Vantage
>    point trees,interval trees, random projection trees, Bregman ball trees,
>    segment trees*. Among these,* I have coded vantage point trees based on
>    a paper cited in the code. You can find the code on my github
>    (https://github.com/parijat10 <https://github.com/parijat10>)* .  As far
>    as I see, X-tree,  R*-tree,  R-tree,  Hilbert-R-tree,  UB-tree,  M-tree,
>    kd tree, cover tree has been implemented in MLPack. Other trees which can
>    be implemented includes space partitioning trees like principal direction
>    trees, k-means trees.*  Please review trees I mentioned, whether it
>    would be fine to implement them. I have basic knowledge of the all the
>    mentioned trees. Please tell if you have some tree type in mind which I
>    have missed or which would be good to work on.*

The ones you have proposed are fine, although the Bregman ball tree will
not work with a lot of mlpack's algorithms, so I might suggest some
other type of tree instead.

>    - Do you want me to write pseudo codes in the proposal?
>    - Also, I was thinking of merging issues similar to issue #275, that is,
>    'rearrange descendants in memory for faster accesses'. As far as I know,
>    only cover tree has this issue. Please correct me any other trees also have
>    this issue, so that I could add this to my project as well.

You're welcome to add #275 to your proposal, but it may already be done
by then.  The RectangleTree class also does not reorder its descendants,
but it is very different that the cover tree and fixing the issue will
require more work.

Thanks,

Ryan

-- 
Ryan Curtin    | "...I still don't know what it means."
ryan at ratml.org |   - Rigby Reardon


More information about the mlpack mailing list