L1 regularization for LogisticRegression?

[Sandy4321]

regarding to regularization for LogisticRegression

https://nbviewer.jupyter.org/github/qinhanmin2014/tiny-sklearn/blob/master/linear_model/LogisticRegression.ipynb

do you have L1 regularization as written in your reference http://fa.bianp.net/blog/2013/numerical-optimizers-for-logistic-regression/

Mathieu • 7 years ago Interesting post! I would mention more clearly that you are focusing on L2-regularized LR, since most solvers considered in this benchmark cannot work for L1-regularized LR.

•Reply•Share › − Avatar Fabian Pedregosa Mod Mathieu • 7 years ago Thanks Mathieu!, I corrected that.

[qinhanmin2014]

Thanks. I don't include L1 penalty because scikit-learn does not support L1 penalty when solver=‘lbfgs’ (and ‘newton-cg’).