That is a good question. I'm not sure what the recommended approach is. I
suppose I can use the binary classification accuracy as my evaluation and
trust the continuos outputs from that are representative. I don't know if
this is good practice or not, but I can't think of another approach since I
don't have continuous ground truth.
On Tue, Jul 2, 2013 at 10:01 PM, Andreas Mueller
<[email protected]>wrote:
> Hey Gene.
> I think it depends on what your loss function will be.
> How do you measure performance for continuous outputs?
>
> Cheers,
> Andy
>
>
> On 07/02/2013 02:40 PM, Gene Kogan wrote:
>
> Yep, thanks allo. I got the same answer mainly in metaoptimize as well. I
> will be using that. Thanks!
>
> best,
> gene
>
>
> On Tue, Jul 2, 2013 at 7:34 PM, Jaques Grobler
> <[email protected]>wrote:
>
>> Didn't see your reply yet, Mathieu :)
>> Thanks
>>
>>
>> 2013/7/2 Jaques Grobler <[email protected]>
>>
>>> Ah when I looked further I see you got some answers here too
>>>
>>>
>>> http://metaoptimize.com/qa/questions/13329/regression-task-trained-on-binary-labels
>>>
>>>
>>>
>>>
>>> 2013/7/2 Jaques Grobler <[email protected]>
>>>
>>>> I would think that Logistic Regression[1] could apply here.. You can
>>>> feed it binary labels and then it will act as a classifier that will return
>>>> for each label the conditional class probability values .
>>>>
>>>> See [2] for scikit-learns implementation
>>>>
>>>> [1] http://en.wikipedia.org/wiki/Logistic_regression
>>>>
>>>> [2]
>>>> http://scikit-learn.org/stable/modules/linear_model.html#logistic-regression
>>>>
>>>> Hope it helps :)
>>>>
>>>>
>>>>
>>>> 2013/7/1 Gene Kogan <[email protected]>
>>>>
>>>>> I have a regression task where I have to assign a continous label
>>>>> between 0 and 1, but my training set contains only binary labels, 0s and
>>>>> 1s. Should I treat this as a classification problem and map the labels to
>>>>> a continous line via some confidence metric (if it's available) or is
>>>>> there
>>>>> a regression algorithm which can be trained on binary labels? What
>>>>> scikits-learn methods will help me achieve this? Thanks!
>>>>>
>>>>> best,
>>>>> gene
>>>>>
>>>>>
>>>>> ------------------------------------------------------------------------------
>>>>> This SF.net email is sponsored by Windows:
>>>>>
>>>>> Build for Windows Store.
>>>>>
>>>>> http://p.sf.net/sfu/windows-dev2dev
>>>>> _______________________________________________
>>>>> Scikit-learn-general mailing list
>>>>> [email protected]
>>>>> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>>>>>
>>>>>
>>>>
>>>
>>
>>
>> ------------------------------------------------------------------------------
>> This SF.net email is sponsored by Windows:
>>
>> Build for Windows Store.
>>
>> http://p.sf.net/sfu/windows-dev2dev
>> _______________________________________________
>> Scikit-learn-general mailing list
>> [email protected]
>> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>>
>>
>
>
> ------------------------------------------------------------------------------
> This SF.net email is sponsored by Windows:
>
> Build for Windows Store.
> http://p.sf.net/sfu/windows-dev2dev
>
>
>
> _______________________________________________
> Scikit-learn-general mailing
> [email protected]https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>
>
>
>
> ------------------------------------------------------------------------------
> This SF.net email is sponsored by Windows:
>
> Build for Windows Store.
>
> http://p.sf.net/sfu/windows-dev2dev
> _______________________________________________
> Scikit-learn-general mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>
>
------------------------------------------------------------------------------
This SF.net email is sponsored by Windows:
Build for Windows Store.
http://p.sf.net/sfu/windows-dev2dev
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
