Implementation of Smooth Bilevel Programming for Sparse Regularization with Scikit-Learn API.
pip install -U git+https://github.com/miclegr/sbp-sr
Requirements are Numpy, Scikit-Learn and Numba
import sbpsr
import sklearn.datasets
n, p = 1000, 20
X, y = sklearn.datasets.make_regression(n,p)
model = sbpsr.Lasso(alpha=1e-3).fit(X,y)
cv = sbpsr.LassoCV().fit(X,y)
Smooth Bilevel Programming for Sparse Regularization
Clarice Poon, Gabriel Peyré