On Thu, Aug 9, 2012 at 3:28 PM, Vlad Niculae <[email protected]> wrote:
> Andy, Mathieu:
> The docs are lacking guidelines and examples on how to tune SVR
> parameters. IIUC, C, gamma, etc should be use just as in SVC. The tricky
> part is epsilon, how should it be set? What are sensible defaults and a
> sensible grid search range?
Hi Vlad,
you could interpret epsilon as the amount of l1 error on the response that
is "tolerated"
by the model.
Sensible values for grid search are in the order of fractions of standard
deviation of the
response. Something along the lines of
epsilons = 2 ** np.arange(-5., 2.) * np.std(y)
Paolo
------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
