On Tue, Mar 14, 2017 at 08:23:20PM +0300, Сидоров Константин wrote:
>>      That's right the HAM module uses a fixed memory size and the
>>      gist looks clean and minimal, however, I would also take a look
>>      at the existing mlpack implementations. If we implement some
>>      tree that we use for the HAM model we have to make sure it's
>>      well tested and that sometimes takes more time than the actual
>>      implementation; if we could reuse some existing code, it's
>>      already tested. But as I said, if it turns out implementing some
>>      specific structure is a better way, we can do that. No need to
>>      use some code that wasn't designed to be used in a totally
>>      different way.
> 
>    I looked in core/tree directory
>    ([2]https://github.com/mlpack/mlpack/tree/master/src/mlpack/core/tree).
>    There are some tree structures, but I can't see a good way to adapt
>    them to HAM architecture - but even if there is, I think it still
>    will be more convenient to implement the gist of my previous
>    letter. The reason is that tree structures from core/tree are
>    created with different problems in mind and maintain
>    problem-specific information in them, which is definitely not going
>    to help because it will hinder both execution and developer time.
>    Maybe I was just unable to find the right tree structure - if so,
>    give me a link to it, because (as you mentioned) having a ready and
>    tested implementation is helpful.

I agree---I don't think mlpack's trees are suited to the HAM situation.
In HAM, my understanding is that a specialized binary tree structure
(built on 1-dimensional data) is used, and the tree is traversed
probabilistically to find the attended/accessed leaf.

The mlpack trees are built for multidimensional data, though, so I am
not sure if there is an easy way to adapt it.  It probably would be
better to either implement a specialized structure or see if something
in std:: would be usable.  But I think because of the way it will be
traversed, it may be somewhat difficult to use some std:: object...
maybe it is possible though.

-- 
Ryan Curtin    | "I'm just going to shoot you once!"
[email protected] |   - Joseph Dunn
_______________________________________________
mlpack mailing list
[email protected]
http://knife.lugatgt.org/cgi-bin/mailman/listinfo/mlpack

Reply via email to