Olivier,
  actually, SGDRegressor is best on boston (of those that give coefs)
so that would be my first choice, for problems big or small.

Grid search ? who has the time ?
OK ... in fact L1-regularization shrinks coefs and R2 towards 0
but av, max |residuals| get worse --

SGDRegressor boston penalty=l1 for alpha in .01 1 2 4
R2 .717  |res| av 2.65  max 15.5  coef [-.2 -1  .5 -1  1  -.7 2.4 ...
R2 .681  |res| av 2.84  max 17.1  coef [-.2 -.7 0  -.2 .1  0  2.1 ...
R2 .628  |res| av 2.95  max 18.8  coef [-.2 -.4 0   0  0   0  1.7 ...
R2 .464  |res| av 3.34  max 22.7  coef [-.3  0  0   0  0   0  .8  ...

Toy 2D yet informative datasets ?
Maybe fun but I think a chimera, and 2D is not nD.
A couple more *real* datasets should be trivial ?

cheers
  -- denis




------------------------------------------------------------------------------
Monitor your physical, virtual and cloud infrastructure from a single
web console. Get in-depth insight into apps, servers, databases, vmware,
SAP, cloud infrastructure, etc. Download 30-day Free Trial.
Pricing starts from $795 for 25 servers or applications!
http://p.sf.net/sfu/zoho_dev2dev_nov
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general

Reply via email to