2013/2/15 Charles-Pierre Astolfi <[email protected]>:
> On Fri, Feb 15, 2013 at 12:53 PM, Olivier Grisel
> <[email protected]> wrote:
>> How do you evaluate which value of alpha is the best if you don't
>> cross validate in one way or another?
>
> Just choose the alpha (from a fixed set) that minimizes the RMSE of
> the prediction of the last time step (or the last n time steps with
> exponential decay). Maybe I'm mistaken but there's no easy way to do
> that with LassoLarsCV.

Indeed there is no way to do that with LassoLarsCV as it does not
support the `sample_weight` fit parameter (to give more importance to
the most recent samples) and does not provide alpha-interpolation
either.

So a new function for interpolation the Lars alpha path would be a
good contribution. It could be either integrated in the existing
lars_path + (Lasso)LarsCV code or as a new `interpolated_lars_path`
function.

-- 
Olivier
http://twitter.com/ogrisel - http://github.com/ogrisel

------------------------------------------------------------------------------
Free Next-Gen Firewall Hardware Offer
Buy your Sophos next-gen firewall before the end March 2013 
and get the hardware for free! Learn more.
http://p.sf.net/sfu/sophos-d2d-feb
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general

Reply via email to