Pushing the docs to dev/ for branch: master, commit 14764061f81b4db311faebe4dc8b2615d8750f29

Author: sklearn-ci

dev/_downloads/auto_examples_jupyter.zip

@@ -15,7 +15,7 @@
       "cell_type": "markdown",
       "metadata": {},
       "source": [
-        "\n# Comparison of Calibration of Classifiers\n\n\nWell calibrated classifiers are probabilistic classifiers for which the output\nof the predict_proba method can be directly interpreted as a confidence level.\nFor instance a well calibrated (binary) classifier should classify the samples\nsuch that among the samples to which it gave a predict_proba value close to\n0.8, approx. 80% actually belong to the positive class.\n\nLogisticRegression returns well calibrated predictions as it directly\noptimizes log-loss. In contrast, the other methods return biased probabilities,\nwith different biases per method:\n\n* GaussianNaiveBayes tends to push probabilities to 0 or 1 (note the counts in\n  the histograms). This is mainly because it makes the assumption that features\n  are conditionally independent given the class, which is not the case in this\n  dataset which contains 2 redundant features.\n\n* RandomForestClassifier shows the opposite behavior: the histograms show\n  peaks at approx. 0.2 and 0.9 probability, while probabilities close to 0 or 1\n  are very rare. An explanation for this is given by Niculescu-Mizil and Caruana\n  [1]: \"Methods such as bagging and random forests that average predictions from\n  a base set of models can have difficulty making predictions near 0 and 1\n  because variance in the underlying base models will bias predictions that\n  should be near zero or one away from these values. Because predictions are\n  restricted to the interval [0,1], errors caused by variance tend to be one-\n  sided near zero and one. For example, if a model should predict p = 0 for a\n  case, the only way bagging can achieve this is if all bagged trees predict\n  zero. If we add noise to the trees that bagging is averaging over, this noise\n  will cause some trees to predict values larger than 0 for this case, thus\n  moving the average prediction of the bagged ensemble away from 0. We observe\n  this effect most strongly with random forests because the base-level trees\n  trained with random forests have relatively high variance due to feature\n  subsetting.\" As a result, the calibration curve shows a characteristic\n  sigmoid shape, indicating that the classifier could trust its \"intuition\"\n  more and return probabilities closer to 0 or 1 typically.\n\n* Support Vector Classification (SVC) shows an even more sigmoid curve as\n  the  RandomForestClassifier, which is typical for maximum-margin methods\n  (compare Niculescu-Mizil and Caruana [1]), which focus on hard samples\n  that are close to the decision boundary (the support vectors).\n\n.. topic:: References:\n\n    .. [1] Predicting Good Probabilities with Supervised Learning,\n          A. Niculescu-Mizil & R. Caruana, ICML 2005\n\n"
+        "\n# Comparison of Calibration of Classifiers\n\n\nWell calibrated classifiers are probabilistic classifiers for which the output\nof the predict_proba method can be directly interpreted as a confidence level.\nFor instance a well calibrated (binary) classifier should classify the samples\nsuch that among the samples to which it gave a predict_proba value close to\n0.8, approx. 80% actually belong to the positive class.\n\nLogisticRegression returns well calibrated predictions as it directly\noptimizes log-loss. In contrast, the other methods return biased probabilities,\nwith different biases per method:\n\n* GaussianNaiveBayes tends to push probabilities to 0 or 1 (note the counts in\n  the histograms). This is mainly because it makes the assumption that features\n  are conditionally independent given the class, which is not the case in this\n  dataset which contains 2 redundant features.\n\n* RandomForestClassifier shows the opposite behavior: the histograms show\n  peaks at approx. 0.2 and 0.9 probability, while probabilities close to 0 or 1\n  are very rare. An explanation for this is given by Niculescu-Mizil and Caruana\n  [1]_: \"Methods such as bagging and random forests that average predictions\n  from a base set of models can have difficulty making predictions near 0 and 1\n  because variance in the underlying base models will bias predictions that\n  should be near zero or one away from these values. Because predictions are\n  restricted to the interval [0,1], errors caused by variance tend to be one-\n  sided near zero and one. For example, if a model should predict p = 0 for a\n  case, the only way bagging can achieve this is if all bagged trees predict\n  zero. If we add noise to the trees that bagging is averaging over, this noise\n  will cause some trees to predict values larger than 0 for this case, thus\n  moving the average prediction of the bagged ensemble away from 0. We observe\n  this effect most strongly with random forests because the base-level trees\n  trained with random forests have relatively high variance due to feature\n  subsetting.\" As a result, the calibration curve shows a characteristic\n  sigmoid shape, indicating that the classifier could trust its \"intuition\"\n  more and return probabilities closer to 0 or 1 typically.\n\n* Support Vector Classification (SVC) shows an even more sigmoid curve as\n  the  RandomForestClassifier, which is typical for maximum-margin methods\n  (compare Niculescu-Mizil and Caruana [1]_), which focus on hard samples\n  that are close to the decision boundary (the support vectors).\n\n.. topic:: References:\n\n    .. [1] Predicting Good Probabilities with Supervised Learning,\n          A. Niculescu-Mizil & R. Caruana, ICML 2005\n\n"
       ]
     },
     {

dev/_downloads/auto_examples_python.zip

@@ -21,8 +21,8 @@
 * RandomForestClassifier shows the opposite behavior: the histograms show
   peaks at approx. 0.2 and 0.9 probability, while probabilities close to 0 or 1
   are very rare. An explanation for this is given by Niculescu-Mizil and Caruana
-  [1]: "Methods such as bagging and random forests that average predictions from
-  a base set of models can have difficulty making predictions near 0 and 1
+  [1]_: "Methods such as bagging and random forests that average predictions
+  from a base set of models can have difficulty making predictions near 0 and 1
   because variance in the underlying base models will bias predictions that
   should be near zero or one away from these values. Because predictions are
   restricted to the interval [0,1], errors caused by variance tend to be one-
@@ -39,7 +39,7 @@
 
 * Support Vector Classification (SVC) shows an even more sigmoid curve as
   the  RandomForestClassifier, which is typical for maximum-margin methods
-  (compare Niculescu-Mizil and Caruana [1]), which focus on hard samples
+  (compare Niculescu-Mizil and Caruana [1]_), which focus on hard samples
   that are close to the decision boundary (the support vectors).
 
 .. topic:: References:

dev/_downloads/plot_compare_calibration.ipynb

@@ -51,7 +51,7 @@
       "cell_type": "markdown",
       "metadata": {},
       "source": [
-        "Compare fit times with and without early stopping\n----------------------------------------------\n\n"
+        "Compare fit times with and without early stopping\n-------------------------------------------------\n\n"
       ]
     },
     {

dev/_downloads/plot_compare_calibration.py

@@ -131,7 +131,7 @@ def autolabel(rects, n_estimators):
 
 #######################################################################
 # Compare fit times with and without early stopping
-# ----------------------------------------------
+# -------------------------------------------------
 
 plt.figure(figsize=(9, 5))
 

dev/_downloads/plot_gradient_boosting_early_stopping.ipynb

@@ -15,7 +15,7 @@
       "cell_type": "markdown",
       "metadata": {},
       "source": [
-        "\n# IsolationForest example\n\n\nAn example using IsolationForest for anomaly detection.\n\nThe IsolationForest 'isolates' observations by randomly selecting a feature\nand then randomly selecting a split value between the maximum and minimum\nvalues of the selected feature.\n\nSince recursive partitioning can be represented by a tree structure, the\nnumber of splittings required to isolate a sample is equivalent to the path\nlength from the root node to the terminating node.\n\nThis path length, averaged over a forest of such random trees, is a measure\nof normality and our decision function.\n\nRandom partitioning produces noticeable shorter paths for anomalies.\nHence, when a forest of random trees collectively produce shorter path lengths\nfor particular samples, they are highly likely to be anomalies.\n\n.. [1] Liu, Fei Tony, Ting, Kai Ming and Zhou, Zhi-Hua. \"Isolation forest.\"\n    Data Mining, 2008. ICDM'08. Eighth IEEE International Conference on.\n\n\n"
+        "\n# IsolationForest example\n\n\nAn example using :class:`sklearn.ensemble.IsolationForest` for anomaly\ndetection.\n\nThe IsolationForest 'isolates' observations by randomly selecting a feature\nand then randomly selecting a split value between the maximum and minimum\nvalues of the selected feature.\n\nSince recursive partitioning can be represented by a tree structure, the\nnumber of splittings required to isolate a sample is equivalent to the path\nlength from the root node to the terminating node.\n\nThis path length, averaged over a forest of such random trees, is a measure\nof normality and our decision function.\n\nRandom partitioning produces noticeable shorter paths for anomalies.\nHence, when a forest of random trees collectively produce shorter path lengths\nfor particular samples, they are highly likely to be anomalies.\n\n\n"
       ]
     },
     {

dev/_downloads/plot_gradient_boosting_early_stopping.py

@@ -3,7 +3,8 @@
 IsolationForest example
 ==========================================
 
-An example using IsolationForest for anomaly detection.
+An example using :class:`sklearn.ensemble.IsolationForest` for anomaly
+detection.
 
 The IsolationForest 'isolates' observations by randomly selecting a feature
 and then randomly selecting a split value between the maximum and minimum
@@ -20,9 +21,6 @@
 Hence, when a forest of random trees collectively produce shorter path lengths
 for particular samples, they are highly likely to be anomalies.
 
-.. [1] Liu, Fei Tony, Ting, Kai Ming and Zhou, Zhi-Hua. "Isolation forest."
-    Data Mining, 2008. ICDM'08. Eighth IEEE International Conference on.
-
 """
 print(__doc__)
 

dev/_downloads/plot_isolation_forest.ipynb

@@ -26,7 +26,7 @@
       },
       "outputs": [],
       "source": [
-        "print(__doc__)\n\nimport numpy as np\nimport matplotlib.pyplot as plt\nfrom sklearn import svm\nfrom sklearn.datasets import make_blobs\n\n# we create two clusters of random points\nn_samples_1 = 1000\nn_samples_2 = 100\ncenters = [[0.0, 0.0], [2.0, 2.0]]\nclusters_std = [1.5, 0.5]\nX, y = make_blobs(n_samples=[n_samples_1, n_samples_2],\n                  centers=centers,\n                  cluster_std=clusters_std,\n                  random_state=0, shuffle=False)\n\n# fit the model and get the separating hyperplane\nclf = svm.SVC(kernel='linear', C=1.0)\nclf.fit(X, y)\n\n# fit the model and get the separating hyperplane using weighted classes\nwclf = svm.SVC(kernel='linear', class_weight={1: 10})\nwclf.fit(X, y)\n\n# plot separating hyperplanes and samples\nplt.scatter(X[:, 0], X[:, 1], c=y, cmap=plt.cm.Paired, edgecolors='k')\nplt.legend()\n\n# plot the decision functions for both classifiers\nax = plt.gca()\nxlim = ax.get_xlim()\nylim = ax.get_ylim()\n\n# create grid to evaluate model\nxx = np.linspace(xlim[0], xlim[1], 30)\nyy = np.linspace(ylim[0], ylim[1], 30)\nYY, XX = np.meshgrid(yy, xx)\nxy = np.vstack([XX.ravel(), YY.ravel()]).T\n\n# get the separating hyperplane\nZ = clf.decision_function(xy).reshape(XX.shape)\n\n# plot decision boundary and margins\na = ax.contour(XX, YY, Z, colors='k', levels=[0], alpha=0.5, linestyles=['-'])\n\n# get the separating hyperplane for weighted classes\nZ = wclf.decision_function(xy).reshape(XX.shape)\n\n# plot decision boundary and margins for weighted classes\nb = ax.contour(XX, YY, Z, colors='r', levels=[0], alpha=0.5, linestyles=['-'])\n\nplt.legend([a.collections[0], b.collections[0]], [\"non weighted\", \"weighted\"],\n           loc=\"upper right\")\nplt.show()"
+        "print(__doc__)\n\nimport numpy as np\nimport matplotlib.pyplot as plt\nfrom sklearn import svm\nfrom sklearn.datasets import make_blobs\n\n# we create two clusters of random points\nn_samples_1 = 1000\nn_samples_2 = 100\ncenters = [[0.0, 0.0], [2.0, 2.0]]\nclusters_std = [1.5, 0.5]\nX, y = make_blobs(n_samples=[n_samples_1, n_samples_2],\n                  centers=centers,\n                  cluster_std=clusters_std,\n                  random_state=0, shuffle=False)\n\n# fit the model and get the separating hyperplane\nclf = svm.SVC(kernel='linear', C=1.0)\nclf.fit(X, y)\n\n# fit the model and get the separating hyperplane using weighted classes\nwclf = svm.SVC(kernel='linear', class_weight={1: 10})\nwclf.fit(X, y)\n\n# plot the samples\nplt.scatter(X[:, 0], X[:, 1], c=y, cmap=plt.cm.Paired, edgecolors='k')\n\n# plot the decision functions for both classifiers\nax = plt.gca()\nxlim = ax.get_xlim()\nylim = ax.get_ylim()\n\n# create grid to evaluate model\nxx = np.linspace(xlim[0], xlim[1], 30)\nyy = np.linspace(ylim[0], ylim[1], 30)\nYY, XX = np.meshgrid(yy, xx)\nxy = np.vstack([XX.ravel(), YY.ravel()]).T\n\n# get the separating hyperplane\nZ = clf.decision_function(xy).reshape(XX.shape)\n\n# plot decision boundary and margins\na = ax.contour(XX, YY, Z, colors='k', levels=[0], alpha=0.5, linestyles=['-'])\n\n# get the separating hyperplane for weighted classes\nZ = wclf.decision_function(xy).reshape(XX.shape)\n\n# plot decision boundary and margins for weighted classes\nb = ax.contour(XX, YY, Z, colors='r', levels=[0], alpha=0.5, linestyles=['-'])\n\nplt.legend([a.collections[0], b.collections[0]], [\"non weighted\", \"weighted\"],\n           loc=\"upper right\")\nplt.show()"
       ]
     }
   ],

dev/_downloads/plot_isolation_forest.py

@@ -49,9 +49,8 @@
 wclf = svm.SVC(kernel='linear', class_weight={1: 10})
 wclf.fit(X, y)
 
-# plot separating hyperplanes and samples
+# plot the samples
 plt.scatter(X[:, 0], X[:, 1], c=y, cmap=plt.cm.Paired, edgecolors='k')
-plt.legend()
 
 # plot the decision functions for both classifiers
 ax = plt.gca()