[SciPy-User] [ANN] scikit.statsmodels 0.2.0 release

josef.pktd@gmai... josef.pktd@gmai...
Thu Feb 18 17:59:14 CST 2010


On Thu, Feb 18, 2010 at 6:43 PM, David Warde-Farley <dwf@cs.toronto.edu> wrote:
> On 18-Feb-10, at 6:28 PM, Robert Kern wrote:
>
>> There's been a recent paper that might be of interest. The associated
>> code is GPLed, alas.
>>
>>  http://www.jstatsoft.org/v33/i01
>>
>> """
>> Abstract:
>> We develop fast algorithms for estimation of generalized linear models
>> with convex penalties. The models include linear regression, two-class
>> logistic regression, and multi- nomial regression problems while the
>> penalties include L1 (the lasso), L2 (ridge regression) and mixtures
>> of the two (the elastic net). The algorithms use cyclical coordinate
>> descent, computed along a regularization path. The methods can handle
>> large problems and can also deal efficiently with sparse features. In
>> comparative timings we find that the new algorithms are considerably
>> faster than competing methods.
>> """
>
> Yep. There's quite a bit of code out there for doing L1 chicanery if
> you're willing to acquiesce to the GPL, also notably this code: http://www.stanford.edu/~boyd/l1_logreg/

but I haven't seen anything yet that would be BSD compatible.

pymvpa is wrapping elastic net and lars with rpy/rpy2 from R. If I
remember then Trevor Hastie, Rob Tibshirani write their code for R.

I think lars is not too difficult, that's one of the reason I have it
in mind, and I have seen some applications in econometrics where the
forecasting performance of lars in combination with PCA was the best
but not lars alone.

Josef




>
> David
> _______________________________________________
> SciPy-User mailing list
> SciPy-User@scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-user
>


More information about the SciPy-User mailing list