Self Organizing map Implementation

["Vaijanath N. Rao" <va...@gmail.com>]

Hi Mahout-developers,

I was looking at Mahout task list, it does not talk about unsupervised 
learning mechanism such as Self-Organizing map. I would like to 
implement Self-Organizing-Map for Mahout ( not sure the map-reduce part 
as I am new to Hadoop), if it needed by the team and does add value to 
Mahout project.

--Thanks and Regards
Vaijanath

Re: Self Organizing map Implementation

[Isabel Drost <ap...@isabel-drost.de>]

On Monday 21 April 2008, Vaijanath N. Rao wrote:
> I think I can get the sequential version well documented and tested by
> end of this week, Would that help?

Sure. That sounds good. It would be nice if you could provide some unit tests 
as well so developers improving your code can be sure they did not break 
anything.

Isabel

-- 
Horse sense is the thing a horse has which keeps it from betting on 
people.		-- W. C. Fields
  |\      _,,,---,,_       Web:   <http://www.isabel-drost.de>
  /,`.-'`'    -.  ;-;;,_
 |,4-  ) )-,_..;\ (  `'-'
'---''(_/--'  `-'\_) (fL)  IM:  <xm...@spaceboyz.net>

Re: Self Organizing map Implementation

["Vaijanath N. Rao" <va...@gmail.com>]

Hi Ted,

Thanks for your reply. I already have sequential reference 
implementation of SOM ( Self-organizing map) and GHSOM (Growing 
Hierarchical SOM) (Not yet fully tested) . I am building application 
around them to demonstrate there usability.

I have already have sequential implementation for Content base Image 
retrieval (Not yet fully tested).  Currently I am writing the 
application for Text similarity, Music similarity and Video similarity.
 
I think I can get the sequential version well documented and tested by 
end of this week, Would that help?

--Thanks and Regards
Vaijanath

Ted Dunning wrote:
> SOM is always interesting to have around.  Other interesting ideas would be
> latent variable representations of some kind.
>
> There are some other unsupervised learning techniques are about to be
> committed (or have recently been committed).  These other methods are mostly
> clustering methods.
>
> We usually start by writing or adapting a sequential reference
> implementation.  Do you think that you could do that?
>
> On 4/20/08 7:26 PM, "Vaijanath N. Rao" <va...@gmail.com> wrote:
>
>   
>> Hi Mahout-developers,
>>
>> I was looking at Mahout task list, it does not talk about unsupervised
>> learning mechanism such as Self-Organizing map. I would like to
>> implement Self-Organizing-Map for Mahout ( not sure the map-reduce part
>> as I am new to Hadoop), if it needed by the team and does add value to
>> Mahout project.
>>
>> --Thanks and Regards
>> Vaijanath
>>
>>     
>
>
>   

Re: Self Organizing map Implementation

[Ted Dunning <td...@veoh.com>]

SOM is always interesting to have around.  Other interesting ideas would be
latent variable representations of some kind.

There are some other unsupervised learning techniques are about to be
committed (or have recently been committed).  These other methods are mostly
clustering methods.

We usually start by writing or adapting a sequential reference
implementation.  Do you think that you could do that?

On 4/20/08 7:26 PM, "Vaijanath N. Rao" <va...@gmail.com> wrote:

> Hi Mahout-developers,
> 
> I was looking at Mahout task list, it does not talk about unsupervised
> learning mechanism such as Self-Organizing map. I would like to
> implement Self-Organizing-Map for Mahout ( not sure the map-reduce part
> as I am new to Hadoop), if it needed by the team and does add value to
> Mahout project.
> 
> --Thanks and Regards
> Vaijanath
>