2013/2/15 Charles-Pierre Astolfi <[email protected]>:
> On Fri, Feb 15, 2013 at 10:35 AM, Gael Varoquaux
> <[email protected]> wrote:
>>> WDYT?
>> I find that lars_path is already a function that is complicated-enough,
>> with a signature difficult to understand, so I'd rather not make it more
>> complex. If there is a need, we can add a similar function that calls
>> lars_path to do what you are suggesting.
>
>
>> However, before we do this, I'd like to understand: what is the usecase
>> and the purpose for this function?
> Which? lars_path or lasso_path or my proposition?
> What I propose in an faster and almost (in the sense that it's not
> computed via sgd and coefs may vary) drop-in replacement for
> lasso_path.

Yes but this is not using the coordinate descent algorithm used by
default in lasso_path hence I find it confusing for the user to use
the lars solver in the lasso_path function. I would rather extend the
lars_path function or create a new one (e.g. interpolated_lars_path)
to be more explicit.

> In fact, some months ago I was trying to do parameter selection for
> Lasso using lasso_path and found it quite slow. I did the
> interpolation by hand from lars_path, and it became very fast, so I
> thought that'd may be worth of a PR.

Sure.

-- 
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