From 7d1689166463fa377e4351e8b9f5cd9030489891 Mon Sep 17 00:00:00 2001
From: Fabian Pedregosa <fabian.pedregosa@inria.fr>
Date: Tue, 14 Dec 2010 19:33:35 +0100
Subject: [PATCH] Add notes on fluctiations of liblinear.
---
scikits/learn/linear_model/logistic.py | 7 +++++++
scikits/learn/linear_model/sparse/logistic.py | 15 +++++++--------
scikits/learn/svm/liblinear.py | 11 +++++++++++
scikits/learn/svm/sparse/liblinear.py | 12 +++++++++++-
4 files changed, 36 insertions(+), 9 deletions(-)
diff --git a/scikits/learn/linear_model/logistic.py b/scikits/learn/linear_model/logistic.py
index a4a18b5584..bd37a2fdaa 100644
--- a/scikits/learn/linear_model/logistic.py
+++ b/scikits/learn/linear_model/logistic.py
@@ -42,6 +42,13 @@ class LogisticRegression(BaseLibLinear, ClassifierMixin):
--------
LinearSVC
+ Notes
+ -----
+ The underlying C implementation uses a random number generator to
+ select features when fitting the model. It is thus not uncommon,
+ to have slightly different results for the same input data. If
+ that happens, try with a smaller eps parameter.
+
References
----------
LIBLINEAR -- A Library for Large Linear Classification
diff --git a/scikits/learn/linear_model/sparse/logistic.py b/scikits/learn/linear_model/sparse/logistic.py
index ca04d7ecd1..0cea75125a 100644
--- a/scikits/learn/linear_model/sparse/logistic.py
+++ b/scikits/learn/linear_model/sparse/logistic.py
@@ -45,18 +45,17 @@ class LogisticRegression(SparseBaseLibLinear, ClassifierMixin):
intercept (a.k.a. bias) added to the decision function.
It is available only when parameter intercept is set to True
- Methods
- -------
- fit(X, Y) : self
- Fit the model
-
- predict(X) : array
- Predict using the model.
-
See also
--------
LinearSVC
+ Notes
+ -----
+ The underlying C implementation uses a random number generator to
+ select features when fitting the model. It is thus not uncommon,
+ to have slightly different results for the same input data. If
+ that happens, try with a smaller eps parameter.
+
References
----------
LIBLINEAR -- A Library for Large Linear Classification
diff --git a/scikits/learn/svm/liblinear.py b/scikits/learn/svm/liblinear.py
index e02be399d4..7bc02de95d 100644
--- a/scikits/learn/svm/liblinear.py
+++ b/scikits/learn/svm/liblinear.py
@@ -43,6 +43,17 @@ class LinearSVC(BaseLibLinear, ClassifierMixin):
`intercept_` : array, shape = [1] if n_classes == 2 else [n_classes]
Constants in decision function.
+ Notes
+ -----
+ The underlying C implementation uses a random number generator to
+ select features when fitting the model. It is thus not uncommon,
+ to have slightly different results for the same input data. If
+ that happens, try with a smaller eps parameter.
+
+ See also
+ --------
+ SVC
+
References
----------
LIBLINEAR -- A Library for Large Linear Classification
diff --git a/scikits/learn/svm/sparse/liblinear.py b/scikits/learn/svm/sparse/liblinear.py
index 57557b7f24..acac283d77 100644
--- a/scikits/learn/svm/sparse/liblinear.py
+++ b/scikits/learn/svm/sparse/liblinear.py
@@ -32,7 +32,6 @@ class LinearSVC(SparseBaseLibLinear, ClassifierMixin):
Select the algorithm to either solve the dual or primal
optimization problem.
-
Attributes
----------
`coef_` : array, shape = [n_features] if n_classes == 2 else [n_classes, n_features]
@@ -42,6 +41,17 @@ class LinearSVC(SparseBaseLibLinear, ClassifierMixin):
`intercept_` : array, shape = [1] if n_classes == 2 else [n_classes]
constants in decision function
+ Notes
+ -----
+ The underlying C implementation uses a random number generator to
+ select features when fitting the model. It is thus not uncommon,
+ to have slightly different results for the same input data. If
+ that happens, try with a smaller eps parameter.
+
+ See also
+ --------
+ SVC
+
References
----------
LIBLINEAR -- A Library for Large Linear Classification
--
GitLab