databionic-esom-user

[Databionic-ESOM-User] Mahalanobis distance function

[Manning, K. <Kev...@ni...>]

Are there any plans to add an implementation of the Mahalanobis distance
calculation or a weighted Mahalanobis calculation? I have an
implementation of the calculation (not using colt or esom) but it is
extremely inefficient for large datasets. I haven't yet been able to get
enough of a handle on the colt or esom API's to try to implement it
myself.

Thanks, Kevin

Re: [Databionic-ESOM-User] Mahalanobis distance function

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

On  0, "Manning, Kevin" <Kev...@ni...> wrote:
> Are there any plans to add an implementation of the Mahalanobis distance
> calculation or a weighted Mahalanobis calculation? I have an
> implementation of the calculation (not using colt or esom) but it is
> extremely inefficient for large datasets. I haven't yet been able to get
> enough of a handle on the colt or esom API's to try to implement it
> myself.

Hi,

it should not be dificult to implement it. Just look at the Euclid
implementation in the esom source.

Greets Niko

Re: [Databionic-ESOM-User] Mahalanobis distance function

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

the distance functions may be a little hard to find.
they are not in the main source code but in the
modified colt library that resides in a separate
directory of the cvs.

implementing your own functions is straightforward.
just add them to colt and extend the command line
handling in the DistanceFactory. for the UI you need
to change the xml file that defines the training dialog.

let us know if you have more questions or want to
commit something.

fabian


Niko Efthymiou wrote:
> On  0, "Manning, Kevin" <Kev...@ni...> wrote:
>   
>> Are there any plans to add an implementation of the Mahalanobis distance
>> calculation or a weighted Mahalanobis calculation? I have an
>> implementation of the calculation (not using colt or esom) but it is
>> extremely inefficient for large datasets. I haven't yet been able to get
>> enough of a handle on the colt or esom API's to try to implement it
>> myself.
>>     
>
> Hi,
>
> it should not be dificult to implement it. Just look at the Euclid
> implementation in the esom source.
>
> Greets Niko
>
> ------------------------------------------------------------------------------
> _______________________________________________
> Databionic-ESOM-User mailing list
> Dat...@li...
> https://lists.sourceforge.net/lists/listinfo/databionic-esom-user
>
>
>