remove normalize parameter anticipating its removal from sklearn (#207)

* remove normalize parameter anticipating its removal from sklearn

* fix warning, increase coverage

* require sklearn>=1.0 in setup

Author: mathurinm

.github/workflows/build.yml

@@ -3,10 +3,10 @@ name: build
 on:
   push:
     branches:
-      - 'master'
+      - 'main'
   pull_request:
     branches:
-      - master
+      - main
 
 jobs:
   build-linux:

celer/dropin_sklearn.py

@@ -19,7 +19,7 @@
 
 
 class Lasso(Lasso_sklearn):
-    """
+    r"""
     Lasso scikit-learn estimator based on Celer solver
 
     The optimization objective for Lasso is::
@@ -61,11 +61,6 @@ class Lasso(Lasso_sklearn):
         Strictly positive weights used in the L1 penalty part of the Lasso
         objective. If None, weights equal to 1 are used.
 
-    normalize : bool, optional (default=False)
-        This parameter is ignored when ``fit_intercept`` is set to False.
-        If True,  the regressors X will be normalized before regression by
-        subtracting the mean and dividing by the l2-norm.
-
     warm_start : bool, optional (default=False)
         When set to True, reuse the solution of the previous call to fit as
         initialization, otherwise, just erase the previous solution.
@@ -113,12 +108,11 @@ class Lasso(Lasso_sklearn):
 
     def __init__(self, alpha=1., max_iter=100, max_epochs=50000, p0=10,
                  verbose=0, tol=1e-4, prune=True, fit_intercept=True,
-                 weights=None, normalize=False, warm_start=False,
+                 weights=None, warm_start=False,
                  positive=False):
         super(Lasso, self).__init__(
             alpha=alpha, tol=tol, max_iter=max_iter,
-            fit_intercept=fit_intercept, normalize=normalize,
-            warm_start=warm_start)
+            fit_intercept=fit_intercept, warm_start=warm_start)
         self.verbose = verbose
         self.max_epochs = max_epochs
         self.p0 = p0
@@ -140,7 +134,7 @@ def path(self, X, y, alphas, coef_init=None, return_n_iter=True, **kwargs):
 
 
 class LassoCV(RegressorMixin, LinearModelCV):
-    """
+    r"""
     LassoCV scikit-learn estimator based on Celer solver
 
     The best model is selected by cross-validation.
@@ -167,11 +161,6 @@ class LassoCV(RegressorMixin, LinearModelCV):
         to false, no intercept will be used in calculations
         (e.g. data is expected to be already centered).
 
-    normalize : bool, optional (default=False)
-        This parameter is ignored when ``fit_intercept`` is set to False.
-        If True,  the regressors X will be normalized before regression by
-        subtracting the mean and dividing by the l2-norm.
-
     max_iter : int, optional
         The maximum number of iterations (subproblem definitions).
 
@@ -243,12 +232,12 @@ class LassoCV(RegressorMixin, LinearModelCV):
     """
 
     def __init__(self, eps=1e-3, n_alphas=100, alphas=None,
-                 fit_intercept=True, normalize=False, max_iter=100,
+                 fit_intercept=True, max_iter=100,
                  tol=1e-4, cv=None, verbose=0, max_epochs=50000, p0=10,
                  prune=True, precompute='auto', positive=False, n_jobs=None):
         super(LassoCV, self).__init__(
             eps=eps, n_alphas=n_alphas, alphas=alphas, max_iter=max_iter,
-            tol=tol, cv=cv, fit_intercept=fit_intercept, normalize=normalize,
+            tol=tol, cv=cv, fit_intercept=fit_intercept,
             verbose=verbose, n_jobs=n_jobs)
         self.max_epochs = max_epochs
         self.p0 = p0
@@ -276,7 +265,7 @@ def _more_tags(self):
 
 
 class MultiTaskLasso(MultiTaskLasso_sklearn):
-    """
+    r"""
     MultiTaskLasso scikit-learn estimator based on Celer solver
 
     The optimization objective for MultiTaskLasso is::
@@ -314,11 +303,6 @@ class MultiTaskLasso(MultiTaskLasso_sklearn):
     fit_intercept : bool, optional (default=True)
         Whether or not to fit an intercept.
 
-    normalize : bool, optional (default=False)
-        This parameter is ignored when ``fit_intercept`` is set to False.
-        If True,  the regressors X will be normalized before regression by
-        subtracting the mean and dividing by the l2-norm.
-
     warm_start : bool, optional (default=False)
         When set to True, reuse the solution of the previous call to fit as
         initialization, otherwise, just erase the previous solution.
@@ -351,11 +335,10 @@ class MultiTaskLasso(MultiTaskLasso_sklearn):
 
     def __init__(self, alpha=1., max_iter=100,
                  max_epochs=50000, p0=10, verbose=0, tol=1e-4, prune=True,
-                 fit_intercept=True, normalize=False, warm_start=False):
+                 fit_intercept=True, warm_start=False):
         super().__init__(
             alpha=alpha, tol=tol, max_iter=max_iter,
-            fit_intercept=fit_intercept, normalize=normalize,
-            warm_start=warm_start)
+            fit_intercept=fit_intercept, warm_start=warm_start)
         self.verbose = verbose
         self.max_epochs = max_epochs
         self.p0 = p0
@@ -382,7 +365,7 @@ def fit(self, X, y):
                              % (n_samples, y.shape[0]))
 
         X, y, X_offset, y_offset, X_scale = _preprocess_data(
-            X, y, self.fit_intercept, self.normalize, copy=False)
+            X, y, self.fit_intercept, copy=False)
 
         if not self.warm_start or not hasattr(self, "coef_"):
             self.coef_ = None
@@ -400,7 +383,7 @@ def fit(self, X, y):
 
 
 class MultiTaskLassoCV(RegressorMixin, LinearModelCV):
-    """
+    r"""
     MultiTaskLassoCV scikit-learn estimator based on Celer solver
 
     The best model is selected by cross-validation.
@@ -427,11 +410,6 @@ class MultiTaskLassoCV(RegressorMixin, LinearModelCV):
         to false, no intercept will be used in calculations
         (e.g. data is expected to be already centered).
 
-    normalize : bool, optional (default=False)
-        This parameter is ignored when ``fit_intercept`` is set to False.
-        If True,  the regressors X will be normalized before regression by
-        subtracting the mean and dividing by the l2-norm.
-
     max_iter : int, optional
         The maximum number of iterations (subproblem definitions).
 
@@ -498,13 +476,13 @@ class MultiTaskLassoCV(RegressorMixin, LinearModelCV):
     """
 
     def __init__(self, eps=1e-3, n_alphas=100, alphas=None,
-                 fit_intercept=True, normalize=False, max_iter=100,
+                 fit_intercept=True, max_iter=100,
                  tol=1e-4, cv=None, verbose=0,
                  max_epochs=50000, p0=10, prune=True, precompute='auto',
                  n_jobs=1):
         super().__init__(
             eps=eps, n_alphas=n_alphas, alphas=alphas, max_iter=max_iter,
-            tol=tol, cv=cv, fit_intercept=fit_intercept, normalize=normalize,
+            tol=tol, cv=cv, fit_intercept=fit_intercept,
             verbose=verbose, n_jobs=n_jobs)
         self.max_epochs = max_epochs
         self.p0 = p0
@@ -531,7 +509,7 @@ def _more_tags(self):
 
 
 class LogisticRegression(LogReg_sklearn):
-    """
+    r"""
     Sparse Logistic regression scikit-learn estimator based on Celer solver.
 
     The optimization objective for sparse Logistic regression is::
@@ -774,11 +752,6 @@ class GroupLasso(Lasso_sklearn):
     fit_intercept : bool, optional (default=True)
         Whether or not to fit an intercept.
 
-    normalize : bool, optional (default=False)
-        This parameter is ignored when ``fit_intercept`` is set to False.
-        If True,  the regressors X will be normalized before regression by
-        subtracting the mean and dividing by the l2-norm.
-
     warm_start : bool, optional (default=False)
         When set to True, reuse the solution of the previous call to fit as
         initialization, otherwise, just erase the previous solution.
@@ -803,7 +776,7 @@ class GroupLasso(Lasso_sklearn):
     >>> clf = GroupLasso(alpha=0.5, groups=[[0, 1], [2]])
     >>> clf.fit([[0, 0, 1], [1, -1, 2], [2, 0, -1]], [1, 1, -1])
     GroupLasso(alpha=0.5, fit_intercept=True,
-    groups=[[0, 1], [2]], max_epochs=50000, max_iter=100, normalize=False,
+    groups=[[0, 1], [2]], max_epochs=50000, max_iter=100,
     p0=10, prune=True, tol=0.0001, verbose=0, warm_start=False)
     >>> print(clf.coef_)
     [-0.         -0.          0.39285714]
@@ -828,10 +801,10 @@ class GroupLasso(Lasso_sklearn):
 
     def __init__(self, groups=1, alpha=1., max_iter=100,
                  max_epochs=50000, p0=10, verbose=0, tol=1e-4, prune=True,
-                 fit_intercept=True, normalize=False, warm_start=False):
+                 fit_intercept=True, warm_start=False):
         super(GroupLasso, self).__init__(
             alpha=alpha, tol=tol, max_iter=max_iter,
-            fit_intercept=fit_intercept, normalize=normalize,
+            fit_intercept=fit_intercept,
             warm_start=warm_start)
         self.groups = groups
         self.verbose = verbose
@@ -854,7 +827,7 @@ def path(self, X, y, alphas, coef_init=None, return_n_iter=True,
 
 
 class GroupLassoCV(LassoCV, LinearModelCV):
-    """
+    r"""
     GroupLassoCV scikit-learn estimator based on Celer solver
 
     The best model is selected by cross-validation.
@@ -892,11 +865,6 @@ class GroupLassoCV(LassoCV, LinearModelCV):
         to false, no intercept will be used in calculations
         (e.g. data is expected to be already centered).
 
-    normalize : bool, optional (default=False)
-        This parameter is ignored when ``fit_intercept`` is set to False.
-        If True,  the regressors X will be normalized before regression by
-        subtracting the mean and dividing by the l2-norm.
-
     max_iter : int, optional
         The maximum number of iterations (subproblem definitions).
 
@@ -968,12 +936,12 @@ class GroupLassoCV(LassoCV, LinearModelCV):
     """
 
     def __init__(self, groups=None, eps=1e-3, n_alphas=100, alphas=None,
-                 fit_intercept=True, normalize=False, max_iter=100,
+                 fit_intercept=True, max_iter=100,
                  tol=1e-4, cv=None, verbose=0, max_epochs=50000, p0=10,
                  prune=True, precompute='auto', positive=False, n_jobs=None):
         super(GroupLassoCV, self).__init__(
             eps=eps, n_alphas=n_alphas, alphas=alphas, max_iter=max_iter,
-            tol=tol, cv=cv, fit_intercept=fit_intercept, normalize=normalize,
+            tol=tol, cv=cv, fit_intercept=fit_intercept,
             verbose=verbose, n_jobs=n_jobs)
         self.groups = groups
         self.max_epochs = max_epochs

celer/tests/test_lasso.py

@@ -115,7 +115,7 @@ def test_Lasso(sparse_X, fit_intercept, positive):
 
     alpha = alpha_max / 2.
     params = dict(alpha=alpha, fit_intercept=fit_intercept, tol=1e-10,
-                  normalize=True, positive=positive)
+                  positive=positive)
     clf = Lasso(**params)
     clf.fit(X, y)
 

celer/tests/test_logreg.py

@@ -46,6 +46,8 @@ def test_LogisticRegression(sparse_X):
     alpha_max = norm(X.T.dot(y), ord=np.inf) / 2
     C = 20. / alpha_max
 
+    clf = LogisticRegression(C=-1)
+    np.testing.assert_raises(ValueError, clf.fit, X, y)
     tol = 1e-8
     clf1 = LogisticRegression(C=C, tol=tol, verbose=0)
     clf1.fit(X, y)

celer/tests/test_mtl.py

@@ -15,22 +15,21 @@
 from celer.utils.testing import build_dataset
 
 
-@pytest.mark.parametrize("sparse_X, fit_intercept, normalize",
-                         itertools.product([0, 1], [0, 1], [0, 1]))
-def test_GroupLasso_Lasso_equivalence(sparse_X, fit_intercept, normalize):
+@pytest.mark.parametrize("sparse_X, fit_intercept",
+                         itertools.product([0, 1], [0, 1]))
+def test_GroupLasso_Lasso_equivalence(sparse_X, fit_intercept):
     """Check that GroupLasso with groups of size 1 gives Lasso."""
     n_features = 1000
     X, y = build_dataset(
         n_samples=100, n_features=n_features, sparse_X=sparse_X)
     alpha_max = norm(X.T @ y, ord=np.inf) / len(y)
     alpha = alpha_max / 10
     clf = Lasso(alpha, tol=1e-12, fit_intercept=fit_intercept,
-                normalize=normalize, verbose=0)
+                verbose=0)
     clf.fit(X, y)
     # take groups of size 1:
     clf1 = GroupLasso(alpha=alpha, groups=1, tol=1e-12,
-                      fit_intercept=fit_intercept, normalize=normalize,
-                      verbose=0)
+                      fit_intercept=fit_intercept, verbose=0)
     clf1.fit(X, y)
 
     np.testing.assert_allclose(clf1.coef_, clf.coef_, atol=1e-6)
@@ -65,12 +64,12 @@ def test_GroupLasso_MultitaskLasso_equivalence():
     np.testing.assert_allclose(alpha_max, other / len(Y_))
 
     alpha = alpha_max / 10
-    clf = MultiTaskLasso(alpha, fit_intercept=False, tol=1e-8, verbose=2)
+    clf = MultiTaskLasso(alpha, fit_intercept=False, tol=1e-8, verbose=0)
     clf.fit(X_, Y_)
 
     groups = [grp.tolist() for grp in grp_indices.reshape(50, 3)]
     clf1 = GroupLasso(alpha=alpha / 3, groups=groups,
-                      fit_intercept=False, tol=1e-8, verbose=2)
+                      fit_intercept=False, tol=1e-8, verbose=0)
     clf1.fit(X, y)
 
     np.testing.assert_allclose(clf1.coef_, clf.coef_.reshape(-1), atol=1e-4)
@@ -110,7 +109,7 @@ def test_MultiTaskLassoCV():
     X, y = build_dataset(n_samples=30, n_features=50, n_targets=3)
 
     params = dict(eps=1e-2, n_alphas=10, tol=1e-12, cv=2, n_jobs=1,
-                  fit_intercept=False, verbose=2)
+                  fit_intercept=False, verbose=0)
 
     clf = MultiTaskLassoCV(**params)
     clf.fit(X, y)
@@ -140,8 +139,7 @@ def test_MultiTaskLasso(fit_intercept):
     alpha_max = np.max(norm(X.T.dot(Y), axis=1)) / X.shape[0]
 
     alpha = alpha_max / 2.
-    params = dict(alpha=alpha, fit_intercept=fit_intercept, tol=1e-10,
-                  normalize=True)
+    params = dict(alpha=alpha, fit_intercept=fit_intercept, tol=1e-10)
     clf = MultiTaskLasso(**params)
     clf.verbose = 2
     clf.fit(X, Y)

setup.py

@@ -35,7 +35,7 @@
       download_url=DOWNLOAD_URL,
       install_requires=['numpy>=1.12', 'seaborn>=0.7', 'scipy>=0.18.0',
                         'matplotlib>=2.0.0', 'Cython>=0.26', 'libsvmdata',
-                        'scikit-learn>=0.24', 'xarray', 'download', 'tqdm'],
+                        'scikit-learn>=1.0', 'xarray', 'download', 'tqdm'],
       packages=find_packages(),
       cmdclass={'build_ext': build_ext},
       ext_modules=[