Table 1. Minimizers employed as solvers in scikit-learn’s logistic regression algorithm.
ElasticNet | No regularization | |||
---|---|---|---|---|
LBFGS1 | ✓ | ✓ | ||
LIBLINEAR (Fan et al., 2008) | ✓ | ✓ | ||
Newton-CG1 | ✓ | ✓ | ||
Newton-Cholesky1 | ✓ | ✓ | ||
SAG (Schmidt, Roux & Bach, 2017) | ✓ | ✓ | ||
SAGA (Defazio, Bach & Lacoste-Julien, 2014) | ✓ | ✓ | ✓ | ✓ |
Note:
See Nocedal & Wright (2006) for further information on these common optimization approaches.