Skip to content

Commit 91ffbff

Browse files
virchanvirchan
authored
DOC fix formatting of intercept_scaling parameter in LogisticRegression-related docs (scikit-learn#31577)
1 parent 0fb2355 commit 91ffbff

File tree

1 file changed

+46
-37
lines changed

1 file changed

+46
-37
lines changed

‎sklearn/linear_model/_logistic.py

Lines changed: 46 additions & 37 deletions
Original file line numberDiff line numberDiff line change
@@ -194,17 +194,19 @@ def _logistic_regression_path(
194194
only supported by the 'saga' solver.
195195
196196
intercept_scaling : float, default=1.
197-
Useful only when the solver 'liblinear' is used
198-
and self.fit_intercept is set to True. In this case, x becomes
199-
[x, self.intercept_scaling],
197+
Useful only when the solver `liblinear` is used
198+
and `self.fit_intercept` is set to `True`. In this case, `x` becomes
199+
`[x, self.intercept_scaling]`,
200200
i.e. a "synthetic" feature with constant value equal to
201-
intercept_scaling is appended to the instance vector.
202-
The intercept becomes ``intercept_scaling * synthetic_feature_weight``.
201+
`intercept_scaling` is appended to the instance vector.
202+
The intercept becomes
203+
``intercept_scaling * synthetic_feature_weight``.
203204
204-
Note! the synthetic feature weight is subject to l1/l2 regularization
205-
as all other features.
206-
To lessen the effect of regularization on synthetic feature weight
207-
(and therefore on the intercept) intercept_scaling has to be increased.
205+
.. note::
206+
The synthetic feature weight is subject to L1 or L2
207+
regularization as all other features.
208+
To lessen the effect of regularization on synthetic feature weight
209+
(and therefore on the intercept) `intercept_scaling` has to be increased.
208210
209211
multi_class : {'ovr', 'multinomial', 'auto'}, default='auto'
210212
If the option chosen is 'ovr', then a binary problem is fit for each
@@ -692,16 +694,19 @@ def _log_reg_scoring_path(
692694
n_samples > n_features.
693695
694696
intercept_scaling : float
695-
Useful only when the solver 'liblinear' is used
696-
and self.fit_intercept is set to True. In this case, x becomes
697-
[x, self.intercept_scaling],
698-
i.e. a "synthetic" feature with constant value equals to
699-
intercept_scaling is appended to the instance vector.
700-
The intercept becomes intercept_scaling * synthetic feature weight
701-
Note! the synthetic feature weight is subject to l1/l2 regularization
702-
as all other features.
703-
To lessen the effect of regularization on synthetic feature weight
704-
(and therefore on the intercept) intercept_scaling has to be increased.
697+
Useful only when the solver `liblinear` is used
698+
and `self.fit_intercept` is set to `True`. In this case, `x` becomes
699+
`[x, self.intercept_scaling]`,
700+
i.e. a "synthetic" feature with constant value equal to
701+
`intercept_scaling` is appended to the instance vector.
702+
The intercept becomes
703+
``intercept_scaling * synthetic_feature_weight``.
704+
705+
.. note::
706+
The synthetic feature weight is subject to L1 or L2
707+
regularization as all other features.
708+
To lessen the effect of regularization on synthetic feature weight
709+
(and therefore on the intercept) `intercept_scaling` has to be increased.
705710
706711
multi_class : {'auto', 'ovr', 'multinomial'}
707712
If the option chosen is 'ovr', then a binary problem is fit for each
@@ -881,17 +886,19 @@ class LogisticRegression(LinearClassifierMixin, SparseCoefMixin, BaseEstimator):
881886
added to the decision function.
882887
883888
intercept_scaling : float, default=1
884-
Useful only when the solver 'liblinear' is used
885-
and self.fit_intercept is set to True. In this case, x becomes
886-
[x, self.intercept_scaling],
889+
Useful only when the solver `liblinear` is used
890+
and `self.fit_intercept` is set to `True`. In this case, `x` becomes
891+
`[x, self.intercept_scaling]`,
887892
i.e. a "synthetic" feature with constant value equal to
888-
intercept_scaling is appended to the instance vector.
889-
The intercept becomes ``intercept_scaling * synthetic_feature_weight``.
893+
`intercept_scaling` is appended to the instance vector.
894+
The intercept becomes
895+
``intercept_scaling * synthetic_feature_weight``.
890896
891-
Note! the synthetic feature weight is subject to l1/l2 regularization
892-
as all other features.
893-
To lessen the effect of regularization on synthetic feature weight
894-
(and therefore on the intercept) intercept_scaling has to be increased.
897+
.. note::
898+
The synthetic feature weight is subject to L1 or L2
899+
regularization as all other features.
900+
To lessen the effect of regularization on synthetic feature weight
901+
(and therefore on the intercept) `intercept_scaling` has to be increased.
895902
896903
class_weight : dict or 'balanced', default=None
897904
Weights associated with classes in the form ``{class_label: weight}``.
@@ -1643,17 +1650,19 @@ class LogisticRegressionCV(LogisticRegression, LinearClassifierMixin, BaseEstima
16431650
best scores across folds are averaged.
16441651
16451652
intercept_scaling : float, default=1
1646-
Useful only when the solver 'liblinear' is used
1647-
and self.fit_intercept is set to True. In this case, x becomes
1648-
[x, self.intercept_scaling],
1653+
Useful only when the solver `liblinear` is used
1654+
and `self.fit_intercept` is set to `True`. In this case, `x` becomes
1655+
`[x, self.intercept_scaling]`,
16491656
i.e. a "synthetic" feature with constant value equal to
1650-
intercept_scaling is appended to the instance vector.
1651-
The intercept becomes ``intercept_scaling * synthetic_feature_weight``.
1657+
`intercept_scaling` is appended to the instance vector.
1658+
The intercept becomes
1659+
``intercept_scaling * synthetic_feature_weight``.
16521660
1653-
Note! the synthetic feature weight is subject to l1/l2 regularization
1654-
as all other features.
1655-
To lessen the effect of regularization on synthetic feature weight
1656-
(and therefore on the intercept) intercept_scaling has to be increased.
1661+
.. note::
1662+
The synthetic feature weight is subject to L1 or L2
1663+
regularization as all other features.
1664+
To lessen the effect of regularization on synthetic feature weight
1665+
(and therefore on the intercept) `intercept_scaling` has to be increased.
16571666
16581667
multi_class : {'auto, 'ovr', 'multinomial'}, default='auto'
16591668
If the option chosen is 'ovr', then a binary problem is fit for each

0 commit comments

Comments
 (0)