Le 06/03/2012 19:19, Andreas Mueller a écrit :
> Hi Adrien.
> Thanks for the offer and thanks for converting people from the dark side ;)
>
> I'm not sure this is the way to go, though.
> There is already quite efficient SGD code in sklearn and this should
> probably
> be extended to handle the multi-class case.
> If you include a separate implementation, there will be a lot of code
> duplication
> and it will probably be non-trivial to get to the speed of the current
> implementation.
I agree with you.

What I had in mind was just to, first, provide a simple, "stand-alone", 
batch implementation of MLR for reference.  Note, that I didn't find any 
in python... Maybe someone else has?

Like you mentioned, this batch version will not scale very well. One 
reason for this is the optimization algorithm used (scipy's BFGS in my 
case).

 From then on, however, it will be easy for the SGD masters to make the 
stochastic version: it will just require re-using the function to 
compute the negative-log-likelihood and its gradient and replace BFGS 
with SGD!

Cheers,

Adrien

>
> On the bright side, after Matthieu's recent contribution it will
> probably not be
> that hard to extend the current code base :)
>
> Cheers,
> Andy
>
>
> On 03/06/2012 07:06 PM, Adrien wrote:
>> Le 06/03/2012 11:25, Andreas a écrit :
>>>>> I agree it would be nice to see a multinomial logistic regreesion in the
>>>>> scikit :).  Any plans for it?
>>>>>
>>>> not on my side but this question comes so often that it might become a
>>>> priority for the project.
>>>>
>>>>
>>> I it a priority for me but fixing API issues and refactoring
>>> is even a higher priority for me ;)
>> I might be able to help on the Multinomial Logistic Regression front.
>>
>> A researcher in our group was telling me this afternoon, that he was
>> considering a slow switch from Matlab to Python. To convince him to
>> leave the dark side, I told him it's fairly easy to convert one's code.
>> He, therefore, sent me his MLR code at the beginning of the afternoon
>> and said: "ok, show me". Challenge accepted! A couple of hours later:
>> done. Please, don't tell my advisor ;-)
>>
>> As you can imagine, the result in such a short time is a bit
>> unsightly... For now, it's very rough and I barely tested in a simple 2D
>> case. I hope I can clean it up a bit, add tests and then make a pull
>> request. I'll let you know about my progress.
>>
>> Cheers,
>>
>> Adrien
>>> ------------------------------------------------------------------------------
>>> Keep Your Developer Skills Current with LearnDevNow!
>>> The most comprehensive online learning library for Microsoft developers
>>> is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
>>> Metro Style Apps, more. Free future releases when you subscribe now!
>>> http://p.sf.net/sfu/learndevnow-d2d
>>> _______________________________________________
>>> Scikit-learn-general mailing list
>>> [email protected]
>>> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>> ------------------------------------------------------------------------------
>> Keep Your Developer Skills Current with LearnDevNow!
>> The most comprehensive online learning library for Microsoft developers
>> is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
>> Metro Style Apps, more. Free future releases when you subscribe now!
>> http://p.sf.net/sfu/learndevnow-d2d
>> _______________________________________________
>> Scikit-learn-general mailing list
>> [email protected]
>> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>
> ------------------------------------------------------------------------------
> Keep Your Developer Skills Current with LearnDevNow!
> The most comprehensive online learning library for Microsoft developers
> is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
> Metro Style Apps, more. Free future releases when you subscribe now!
> http://p.sf.net/sfu/learndevnow-d2d
> _______________________________________________
> Scikit-learn-general mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general


------------------------------------------------------------------------------
Keep Your Developer Skills Current with LearnDevNow!
The most comprehensive online learning library for Microsoft developers
is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
Metro Style Apps, more. Free future releases when you subscribe now!
http://p.sf.net/sfu/learndevnow-d2d
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general

Reply via email to