Python wrapper around glmnet R package for generalized linear models, providing scikit-learn style estimators.
The glmnet R package was written by the creators of the lasso and elastic net. It is highly optimized and tends to outdo other implementations in both model performance and speed.
Currently supports basic linear, logistic, multinomial (multi-class), and Poisson regression models.
To install dependencies:
conda install numpy pandas scikit-learn rpy2=2.9.4
Rscript -e 'install.packages("glmnet", repos="https://cloud.r-project.org")'
On some systems, it may be necessary to run
conda install -c r libiconv
prior to the Rscript command.
Recent Rpy2 versions have problems with some type conversions.