chung wrote:
> Dear Nina,
>
> On Sun, 2009-11-01 at 11:23 +0200, Nina Jeliazkova wrote:
>   
>> Dear Pantelis,
>>
>>
>>     PMML 4.0 (June 2009) Extensions of Existing Elements: Addition of
>>     multi-class classification for Support Vector Machines
>>     <http://en.wikipedia.org/wiki/Support_Vector_Machines>,
>>
>>
>> It seems to be specific to SVM, rather than generic support for multi
>> class models, but this needs to be verified.  Regarding non-PMML API, if
>> the number of target variables is not be fixed to 1, will this be
>> sufficient?
>>
>>     
>
> In that case, we should have 2 different models. One for each target
> feature.
>   
Not necessary, there are class of algorithms that can predict multiple
targets at once.

Regards,
Nina
>   
>> Best regards,
>> Nina
>>
>> chung wrote:
>>     
>>> Dear All,
>>>   Is it possible to represent a classification model in the standard
>>> PMML format if it consists of more than one discrimination functions
>>> (i.e. there more than one classes)? If not we have to think about how we
>>> could represent multi-class models. Intuitively, I say we could provide
>>> a model for every discrimination function or modify the PMML schema but
>>> I'm not much in favour of this solution.
>>>   Well, I'm looking forward to read your thoughts on that.
>>>
>>> Best Regards,
>>> Pantelis
>>>
>>> _______________________________________________
>>> Development mailing list
>>> Development at opentox.org
>>> http://www.opentox.org/mailman/listinfo/development
>>>   
>>>       
>> _______________________________________________
>> Development mailing list
>> Development at opentox.org
>> http://www.opentox.org/mailman/listinfo/development
>>
>>     
>
> _______________________________________________
> Development mailing list
> Development at opentox.org
> http://www.opentox.org/mailman/listinfo/development
>