On 12/30/2011 10:15 AM, Gael Varoquaux wrote:
> On Fri, Dec 30, 2011 at 10:09:59AM +0100, Olivier Grisel wrote:
>>>   * You could use the decision function, (decision_function method of the
>>>    LinearSVC) although this is clearly a hack.
>> Why is this a hack? ROC is only concerned with the relative positions
>> of the decision threshold, not the probability normalization AFAIK, am
>> I wrong?
> You are right, but there is nothing that tell us that the decision
> function is really related to a success probability. That said, the Platt
> method used to implement predict_proba relies on a similar hack.
>
It might be that I haven't really understood the meaning of ROC
curves, but I thought it worked like @ogrisel said.
Whatever the correct method to produce a ROC curve
from a linear classifier, I'm pretty sure that using the decision
function is very common in the ML literature.

Cheers,
Andy


------------------------------------------------------------------------------
Ridiculously easy VDI. With Citrix VDI-in-a-Box, you don't need a complex
infrastructure or vast IT resources to deliver seamless, secure access to
virtual desktops. With this all-in-one solution, easily deploy virtual 
desktops for less than the cost of PCs and save 60% on VDI infrastructure 
costs. Try it free! http://p.sf.net/sfu/Citrix-VDIinabox
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general

Reply via email to