MNT update Travis dependencies to latest available versions (scikit-l…

…earn#11714)

.travis.yml

@@ -38,13 +38,13 @@ matrix:
  NUMPY_VERSION="1.10.4" SCIPY_VERSION="0.16.1" CYTHON_VERSION="0.25.2"
  PILLOW_VERSION="4.0.0" COVERAGE=true
  if: type != cron
- # This environment tests the newest supported Anaconda release.
+ # This environment tests the latest available dependencies.
  # It runs tests requiring pandas and PyAMG.
  # It also runs with the site joblib instead of the vendored copy of joblib.
- - env: DISTRIB="conda" PYTHON_VERSION="3.6.2" INSTALL_MKL="true"
- NUMPY_VERSION="1.14.2" SCIPY_VERSION="1.0.0" PANDAS_VERSION="0.20.3"
- CYTHON_VERSION="0.26.1" PYAMG_VERSION="3.3.2" PILLOW_VERSION="4.3.0"
- JOBLIB_VERSION="0.12" COVERAGE=true
+ - env: DISTRIB="conda" PYTHON_VERSION="*" INSTALL_MKL="true"
+ NUMPY_VERSION="*" SCIPY_VERSION="*" PANDAS_VERSION="*"
+ CYTHON_VERSION="*" PYAMG_VERSION="*" PILLOW_VERSION="*"
+ JOBLIB_VERSION="*" COVERAGE=true
  CHECK_PYTEST_SOFT_DEPENDENCY="true" TEST_DOCSTRINGS="true"
  SKLEARN_SITE_JOBLIB=1
  if: type != cron

examples/applications/plot_prediction_latency.py

@@ -26,7 +26,6 @@
 
 from sklearn.preprocessing import StandardScaler
 from sklearn.model_selection import train_test_split
-from scipy.stats import scoreatpercentile
 from sklearn.datasets.samples_generator import make_regression
 from sklearn.ensemble.forest import RandomForestRegressor
 from sklearn.linear_model.ridge import Ridge
@@ -50,7 +49,7 @@ def atomic_benchmark_estimator(estimator, X_test, verbose=False):
  estimator.predict(instance)
  runtimes[i] = time.time() - start
  if verbose:
- print("atomic_benchmark runtimes:", min(runtimes), scoreatpercentile(
+ print("atomic_benchmark runtimes:", min(runtimes), np.percentile(
  runtimes, 50), max(runtimes))
  return runtimes
 
@@ -65,7 +64,7 @@ def bulk_benchmark_estimator(estimator, X_test, n_bulk_repeats, verbose):
  runtimes[i] = time.time() - start
  runtimes = np.array(list(map(lambda x: x / float(n_instances), runtimes)))
  if verbose:
- print("bulk_benchmark runtimes:", min(runtimes), scoreatpercentile(
+ print("bulk_benchmark runtimes:", min(runtimes), np.percentile(
  runtimes, 50), max(runtimes))
  return runtimes
 
@@ -207,8 +206,8 @@ def n_feature_influence(estimators, n_train, n_test, n_features, percentile):
  estimator.fit(X_train, y_train)
  gc.collect()
  runtimes = bulk_benchmark_estimator(estimator, X_test, 30, False)
- percentiles[cls_name][n] = 1e6 * scoreatpercentile(runtimes,
-  percentile)
+ percentiles[cls_name][n] = 1e6 * np.percentile(runtimes,
+ percentile)
  return percentiles
 
 

sklearn/covariance/elliptic_envelope.py

@@ -106,8 +106,7 @@ def fit(self, X, y=None):
  y : (ignored)
  """
  super(EllipticEnvelope, self).fit(X)
- self.offset_ = sp.stats.scoreatpercentile(
- -self.dist_, 100. * self.contamination)
+ self.offset_ = np.percentile(-self.dist_, 100. * self.contamination)
  return self
 
  def decision_function(self, X, raw_values=None):

sklearn/ensemble/gradient_boosting.py

@@ -38,7 +38,6 @@
 import numbers
 import numpy as np
 
-from scipy import stats
 from scipy.sparse import csc_matrix
 from scipy.sparse import csr_matrix
 from scipy.sparse import issparse
@@ -91,7 +90,7 @@ def fit(self, X, y, sample_weight=None):
  Individual weights for each sample
  """
  if sample_weight is None:
- self.quantile = stats.scoreatpercentile(y, self.alpha * 100.0)
+ self.quantile = np.percentile(y, self.alpha * 100.0)
  else:
  self.quantile = _weighted_percentile(y, sample_weight,
  self.alpha * 100.0)
@@ -608,7 +607,7 @@ def __call__(self, y, pred, sample_weight=None):
  gamma = self.gamma
  if gamma is None:
  if sample_weight is None:
- gamma = stats.scoreatpercentile(np.abs(diff), self.alpha * 100)
+ gamma = np.percentile(np.abs(diff), self.alpha * 100)
  else:
  gamma = _weighted_percentile(np.abs(diff), sample_weight, self.alpha * 100)
 
@@ -641,7 +640,7 @@ def negative_gradient(self, y, pred, sample_weight=None, **kargs):
  pred = pred.ravel()
  diff = y - pred
  if sample_weight is None:
- gamma = stats.scoreatpercentile(np.abs(diff), self.alpha * 100)
+ gamma = np.percentile(np.abs(diff), self.alpha * 100)
  else:
  gamma = _weighted_percentile(np.abs(diff), sample_weight, self.alpha * 100)
  gamma_mask = np.abs(diff) <= gamma

sklearn/ensemble/iforest.py

@@ -267,8 +267,8 @@ def fit(self, X, y=None, sample_weight=None):
  "'auto' when behaviour == 'old'.")
 
  self.offset_ = -0.5
- self._threshold_ = sp.stats.scoreatpercentile(
- self.decision_function(X), 100. * self._contamination)
+ self._threshold_ = np.percentile(self.decision_function(X),
+   100. * self._contamination)
 
  return self
 
@@ -281,8 +281,8 @@ def fit(self, X, y=None, sample_weight=None):
 
  # else, define offset_ wrt contamination parameter, so that the
  # threshold_ attribute is implicitly 0 and is not needed anymore:
- self.offset_ = sp.stats.scoreatpercentile(
- self.score_samples(X), 100. * self._contamination)
+ self.offset_ = np.percentile(self.score_samples(X),
+   100. * self._contamination)
 
  return self
 

sklearn/feature_selection/univariate_selection.py

@@ -438,8 +438,7 @@ def _get_support_mask(self):
  return np.zeros(len(self.scores_), dtype=np.bool)
 
  scores = _clean_nans(self.scores_)
- threshold = stats.scoreatpercentile(scores,
- 100 - self.percentile)
+ threshold = np.percentile(scores, 100 - self.percentile)
  mask = scores > threshold
  ties = np.where(scores == threshold)[0]
  if len(ties):

sklearn/neighbors/lof.py

@@ -4,7 +4,6 @@
 
 import numpy as np
 import warnings
-from scipy.stats import scoreatpercentile
 
 from .base import NeighborsBase
 from .base import KNeighborsMixin
@@ -262,8 +261,8 @@ def fit(self, X, y=None):
  # inliers score around -1 (the higher, the less abnormal).
  self.offset_ = -1.5
  else:
- self.offset_ = scoreatpercentile(
- self.negative_outlier_factor_, 100. * self._contamination)
+ self.offset_ = np.percentile(self.negative_outlier_factor_,
+   100. * self._contamination)
 
  return self