When using the LogisticRegression
module of scikit-learn, I do not understand the correspondence between Penalty and Solver Official document I looked it up, but it wasn't ready to tell at a glance, and I was confused for a while. .. .. Therefore, first, I will summarize the correspondence / non-correspondence relationship between Penalty and Solver in a list, and I will add detailed information on Solver and Penalty in the future.
Penalty / Solver | newton-cg | lbfgs(default) | liblinear | sag | saga |
---|---|---|---|---|---|
L1 | × | × | ○ | × | ○ |
L2(default) | ○ | ○ | × | ○ | ○ |
Elasticnet | × | × | × | × | ○ |
None | ○ | ○ | × | ○ | ○ |