databionic-esom-user

[Databionic-ESOM-User] Bray-Curtis distances

[rimet f. <rim...@ya...>]

Hello,
I've tried your eSOM program on ecological data (phytoplancton counts of a french lake). It produces interesting maps and groups. Nevertheless I wanted to know if you plan to add in the data space distance function other distances more adapted to ecological data, as for instance Sorensen (Bray & Curtis) Distance or Jaccard Distance ?
Thanks for your answer,
Frédéric RIMET

__________________________________________________
Do You Yahoo!?
En finir avec le spam? Yahoo! Mail vous offre la meilleure protection possible contre les messages non sollicités 
http://mail.yahoo.fr Yahoo! Mail 

Re: [Databionic-ESOM-User] Bray-Curtis distances

[Niko E. <ne...@Ma...>]

On Wednesday 14 May 2008, rimet frederic wrote:
> Hello,
> I've tried your eSOM program on ecological data (phytoplancton counts
> of a french lake). It produces interesting maps and groups.
> Nevertheless I wanted to know if you plan to add in the data space
> distance function other distances more adapted to ecological data, as
> for instance Sorensen (Bray & Curtis) Distance or Jaccard Distance ?
> Thanks for your answer,
> Frédéric RIMET

Hello,

as far as I know there are no such plans. However adding this metrics 
should not be wery hard. You just have to implement them as 
	
	cern.colt.function.ThresholdVectorVectorFunction
 	cern.colt.function.VectorVectorFunction

objects and add them to 
	
	databionics.math.DistanceFactory
	databionics.esom.Train

and that schould be about it (my crystal ball reads ~100 lines of code). 
Please consider submiting a patch if you do add them.


Greats Niko

Re: [Databionic-ESOM-User] Bray-Curtis distances

[Fabian M. <fa...@my...>]

Frederic,

Bray Curtis could be added very easily as it is already implemented in
colt.

The Jaccard distance measures the difference between sets whereas the
ESOM works with numeric vectors. what does your input look like? do you
have categoric vectors represented as real numbers?

Please note that for a meaningful ESOM training you need a distance
function for your vectors and the (weighted) mean of you input vectors
must represent an 'average' of the examples. This might not be the case
for categoric data depending on the encoding. A binary encoding might
work.

A little while ago I added the hamming distance per special request.
Hamming is defined for binary vectors. The ESOM training generated
vectors using the default weighted mean but the distance function
rounds each entry to the nearest integer and counts the number of
identical entries in the two vectors.

best
fabian

Niko Efthymiou wrote:
> On Wednesday 14 May 2008, rimet frederic wrote:
>   
>> Hello,
>> I've tried your eSOM program on ecological data (phytoplancton counts
>> of a french lake). It produces interesting maps and groups.
>> Nevertheless I wanted to know if you plan to add in the data space
>> distance function other distances more adapted to ecological data, as
>> for instance Sorensen (Bray & Curtis) Distance or Jaccard Distance ?
>> Thanks for your answer,
>> Frédéric RIMET
>>     
>
> Hello,
>
> as far as I know there are no such plans. However adding this metrics 
> should not be wery hard. You just have to implement them as 
> 	
> 	cern.colt.function.ThresholdVectorVectorFunction
>  	cern.colt.function.VectorVectorFunction
>
> objects and add them to 
> 	
> 	databionics.math.DistanceFactory
> 	databionics.esom.Train
>
> and that schould be about it (my crystal ball reads ~100 lines of code). 
> Please consider submiting a patch if you do add them.
>
>
> Greats Niko
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by: Microsoft 
> Defy all challenges. Microsoft(R) Visual Studio 2008. 
> http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
> _______________________________________________
> Databionic-ESOM-User mailing list
> Dat...@li...
> https://lists.sourceforge.net/lists/listinfo/databionic-esom-user
>
>
>