Implementation:Scikit learn Scikit learn LogisticRegression Init

Field	Value
source	scikit-learn\|https://github.com/scikit-learn/scikit-learn
domains	Data_Science, Machine_Learning
last_updated	2026-02-08 15:00 GMT

Overview

Concrete tool for instantiating a logistic regression classifier provided by scikit-learn.

Description

The LogisticRegression.__init__ method configures a logistic regression estimator with all hyperparameters required for subsequent training. Logistic regression is a linear model for classification that models the posterior probability of each class using the logistic (sigmoid) function. The constructor accepts parameters controlling regularization, optimization, convergence, and parallelism without performing any computation.

Usage

Setting up a baseline linear classifier for binary or multiclass classification tasks.
Configuring regularization strength (C) and type (l1_ratio) to control model complexity.
Selecting an optimization solver (solver) appropriate for the dataset size and regularization type.
Preparing an estimator instance for use in pipelines, grid search, or standalone training.

Code Reference

Source Location

sklearn/linear_model/_logistic.py, method LogisticRegression.__init__

Signature

def __init__(
    self,
    penalty="deprecated",
    *,
    C=1.0,
    l1_ratio=0.0,
    dual=False,
    tol=1e-4,
    fit_intercept=True,
    intercept_scaling=1,
    class_weight=None,
    random_state=None,
    solver="lbfgs",
    max_iter=100,
    verbose=0,
    warm_start=False,
    n_jobs=None,
):

Import

from sklearn.linear_model import LogisticRegression

I/O Contract

Inputs (Constructor Parameters)

Parameter	Type	Default	Description
`penalty`	str	`"deprecated"`	Regularization norm. Deprecated in v1.8; use `l1_ratio` and `C` instead.
`C`	float	`1.0`	Inverse of regularization strength. Smaller values specify stronger regularization. Set to `np.inf` to disable regularization.
`l1_ratio`	float	`0.0`	The Elastic-Net mixing parameter. `0.0` is equivalent to L2, `1.0` is L1, and values in between produce Elastic-Net regularization.
`dual`	bool	`False`	Dual formulation (only for L2 penalty with liblinear solver). Prefer `dual=False` when n_samples > n_features.
`tol`	float	`1e-4`	Tolerance for the stopping criterion of the solver.
`fit_intercept`	bool	`True`	Whether to add a constant (bias) term to the decision function.
`intercept_scaling`	float	`1`	Scaling factor for the intercept term when using the liblinear solver.
`class_weight`	dict or "balanced" or None	`None`	Weights associated with classes. If `"balanced"`, weights are inversely proportional to class frequencies.
`random_state`	int, RandomState, or None	`None`	Seed for reproducibility. Used when `solver` is `"sag"`, `"saga"`, or `"liblinear"`.
`solver`	str	`"lbfgs"`	Optimization algorithm. Choices: `"lbfgs"`, `"liblinear"`, `"newton-cg"`, `"newton-cholesky"`, `"sag"`, `"saga"`.
`max_iter`	int	`100`	Maximum number of iterations for the solver to converge.
`verbose`	int	`0`	Verbosity level for solver output.
`warm_start`	bool	`False`	If `True`, reuse the solution of the previous call to `fit` as initialization.
`n_jobs`	int or None	`None`	Deprecated since v1.8. Previously controlled parallel jobs for OvR multiclass fitting.

Outputs

Return	Type	Description
instance	`LogisticRegression`	A configured (unfitted) estimator instance with all hyperparameters stored as attributes.

Usage Examples

Default instantiation:

from sklearn.linear_model import LogisticRegression

clf = LogisticRegression()
print(clf.get_params())
# {'C': 1.0, 'class_weight': None, 'dual': False, 'fit_intercept': True, ...}

Custom hyperparameters:

from sklearn.linear_model import LogisticRegression

clf = LogisticRegression(
    C=0.5,
    l1_ratio=0.0,
    solver="lbfgs",
    max_iter=200,
    random_state=42,
)

Using L1 regularization with the saga solver:

from sklearn.linear_model import LogisticRegression

clf = LogisticRegression(
    C=1.0,
    l1_ratio=1.0,
    solver="saga",
    max_iter=500,
)

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment