Pushing the docs to dev/ for branch: main, commit a744c476511e8d87952be7910143a764d7152ef4

Author: sklearn-ci

dev/_downloads/02192f99342d6d1323161978b6c80bfc/plot_ols_ridge.zip

@@ -88,12 +88,12 @@ def fit_predict(estimator, X):
     tic = perf_counter()
     if estimator[-1].__class__.__name__ == "LocalOutlierFactor":
         estimator.fit(X)
-        y_pred = estimator[-1].negative_outlier_factor_
+        y_score = estimator[-1].negative_outlier_factor_
     else:  # "IsolationForest"
-        y_pred = estimator.fit(X).decision_function(X)
+        y_score = estimator.fit(X).decision_function(X)
     toc = perf_counter()
     print(f"Duration for {model_name}: {toc - tic:.2f} s")
-    return y_pred
+    return y_score
 
 
 # %%
@@ -138,7 +138,7 @@ def fit_predict(estimator, X):
 
 # %%
 y_true = {}
-y_pred = {"LOF": {}, "IForest": {}}
+y_score = {"LOF": {}, "IForest": {}}
 model_names = ["LOF", "IForest"]
 cat_columns = ["protocol_type", "service", "flag"]
 
@@ -150,7 +150,7 @@ def fit_predict(estimator, X):
         lof_kw={"n_neighbors": int(n_samples * anomaly_frac)},
         iforest_kw={"random_state": 42},
     )
-    y_pred[model_name]["KDDCup99 - SA"] = fit_predict(model, X)
+    y_score[model_name]["KDDCup99 - SA"] = fit_predict(model, X)
 
 # %%
 # Forest covertypes dataset
@@ -185,7 +185,7 @@ def fit_predict(estimator, X):
         lof_kw={"n_neighbors": int(n_samples * anomaly_frac)},
         iforest_kw={"random_state": 42},
     )
-    y_pred[model_name]["forestcover"] = fit_predict(model, X)
+    y_score[model_name]["forestcover"] = fit_predict(model, X)
 
 # %%
 # Ames Housing dataset
@@ -242,7 +242,7 @@ def fit_predict(estimator, X):
         lof_kw={"n_neighbors": int(n_samples * anomaly_frac)},
         iforest_kw={"random_state": 42},
     )
-    y_pred[model_name]["ames_housing"] = fit_predict(model, X)
+    y_score[model_name]["ames_housing"] = fit_predict(model, X)
 
 # %%
 # Cardiotocography dataset
@@ -271,7 +271,7 @@ def fit_predict(estimator, X):
         lof_kw={"n_neighbors": int(n_samples * anomaly_frac)},
         iforest_kw={"random_state": 42},
     )
-    y_pred[model_name]["cardiotocography"] = fit_predict(model, X)
+    y_score[model_name]["cardiotocography"] = fit_predict(model, X)
 
 # %%
 # Plot and interpret results
@@ -299,7 +299,7 @@ def fit_predict(estimator, X):
     for model_idx, model_name in enumerate(model_names):
         display = RocCurveDisplay.from_predictions(
             y_true[dataset_name],
-            y_pred[model_name][dataset_name],
+            y_score[model_name][dataset_name],
             pos_label=pos_label,
             name=model_name,
             ax=ax,
@@ -346,10 +346,10 @@ def fit_predict(estimator, X):
 for model_idx, (linestyle, n_neighbors) in enumerate(zip(linestyles, n_neighbors_list)):
     model.set_params(localoutlierfactor__n_neighbors=n_neighbors)
     model.fit(X)
-    y_pred = model[-1].negative_outlier_factor_
+    y_score = model[-1].negative_outlier_factor_
     display = RocCurveDisplay.from_predictions(
         y,
-        y_pred,
+        y_score,
         pos_label=pos_label,
         name=f"n_neighbors = {n_neighbors}",
         ax=ax,
@@ -386,10 +386,10 @@ def fit_predict(estimator, X):
 ):
     model = make_pipeline(preprocessor, lof)
     model.fit(X)
-    y_pred = model[-1].negative_outlier_factor_
+    y_score = model[-1].negative_outlier_factor_
     display = RocCurveDisplay.from_predictions(
         y,
-        y_pred,
+        y_score,
         pos_label=pos_label,
         name=str(preprocessor).split("(")[0],
         ax=ax,
@@ -438,10 +438,10 @@ def fit_predict(estimator, X):
 ):
     model = make_pipeline(preprocessor, lof)
     model.fit(X)
-    y_pred = model[-1].negative_outlier_factor_
+    y_score = model[-1].negative_outlier_factor_
     display = RocCurveDisplay.from_predictions(
         y,
-        y_pred,
+        y_score,
         pos_label=pos_label,
         name=str(preprocessor).split("(")[0],
         ax=ax,

dev/_downloads/02fe21a1f5d14bd1ebbe34aa905d9bf6/plot_multilabel.zip

@@ -51,7 +51,7 @@
       },
       "outputs": [],
       "source": [
-        "from time import perf_counter\n\n\ndef fit_predict(estimator, X):\n    tic = perf_counter()\n    if estimator[-1].__class__.__name__ == \"LocalOutlierFactor\":\n        estimator.fit(X)\n        y_pred = estimator[-1].negative_outlier_factor_\n    else:  # \"IsolationForest\"\n        y_pred = estimator.fit(X).decision_function(X)\n    toc = perf_counter()\n    print(f\"Duration for {model_name}: {toc - tic:.2f} s\")\n    return y_pred"
+        "from time import perf_counter\n\n\ndef fit_predict(estimator, X):\n    tic = perf_counter()\n    if estimator[-1].__class__.__name__ == \"LocalOutlierFactor\":\n        estimator.fit(X)\n        y_score = estimator[-1].negative_outlier_factor_\n    else:  # \"IsolationForest\"\n        y_score = estimator.fit(X).decision_function(X)\n    toc = perf_counter()\n    print(f\"Duration for {model_name}: {toc - tic:.2f} s\")\n    return y_score"
       ]
     },
     {
@@ -87,7 +87,7 @@
       },
       "outputs": [],
       "source": [
-        "y_true = {}\ny_pred = {\"LOF\": {}, \"IForest\": {}}\nmodel_names = [\"LOF\", \"IForest\"]\ncat_columns = [\"protocol_type\", \"service\", \"flag\"]\n\ny_true[\"KDDCup99 - SA\"] = y\nfor model_name in model_names:\n    model = make_estimator(\n        name=model_name,\n        categorical_columns=cat_columns,\n        lof_kw={\"n_neighbors\": int(n_samples * anomaly_frac)},\n        iforest_kw={\"random_state\": 42},\n    )\n    y_pred[model_name][\"KDDCup99 - SA\"] = fit_predict(model, X)"
+        "y_true = {}\ny_score = {\"LOF\": {}, \"IForest\": {}}\nmodel_names = [\"LOF\", \"IForest\"]\ncat_columns = [\"protocol_type\", \"service\", \"flag\"]\n\ny_true[\"KDDCup99 - SA\"] = y\nfor model_name in model_names:\n    model = make_estimator(\n        name=model_name,\n        categorical_columns=cat_columns,\n        lof_kw={\"n_neighbors\": int(n_samples * anomaly_frac)},\n        iforest_kw={\"random_state\": 42},\n    )\n    y_score[model_name][\"KDDCup99 - SA\"] = fit_predict(model, X)"
       ]
     },
     {
@@ -116,7 +116,7 @@
       },
       "outputs": [],
       "source": [
-        "y_true[\"forestcover\"] = y\nfor model_name in model_names:\n    model = make_estimator(\n        name=model_name,\n        lof_kw={\"n_neighbors\": int(n_samples * anomaly_frac)},\n        iforest_kw={\"random_state\": 42},\n    )\n    y_pred[model_name][\"forestcover\"] = fit_predict(model, X)"
+        "y_true[\"forestcover\"] = y\nfor model_name in model_names:\n    model = make_estimator(\n        name=model_name,\n        lof_kw={\"n_neighbors\": int(n_samples * anomaly_frac)},\n        iforest_kw={\"random_state\": 42},\n    )\n    y_score[model_name][\"forestcover\"] = fit_predict(model, X)"
       ]
     },
     {
@@ -163,7 +163,7 @@
       },
       "outputs": [],
       "source": [
-        "from sklearn.compose import make_column_selector as selector\n\ncategorical_columns_selector = selector(dtype_include=\"category\")\ncat_columns = categorical_columns_selector(X)\n\ny_true[\"ames_housing\"] = y\nfor model_name in model_names:\n    model = make_estimator(\n        name=model_name,\n        categorical_columns=cat_columns,\n        lof_kw={\"n_neighbors\": int(n_samples * anomaly_frac)},\n        iforest_kw={\"random_state\": 42},\n    )\n    y_pred[model_name][\"ames_housing\"] = fit_predict(model, X)"
+        "from sklearn.compose import make_column_selector as selector\n\ncategorical_columns_selector = selector(dtype_include=\"category\")\ncat_columns = categorical_columns_selector(X)\n\ny_true[\"ames_housing\"] = y\nfor model_name in model_names:\n    model = make_estimator(\n        name=model_name,\n        categorical_columns=cat_columns,\n        lof_kw={\"n_neighbors\": int(n_samples * anomaly_frac)},\n        iforest_kw={\"random_state\": 42},\n    )\n    y_score[model_name][\"ames_housing\"] = fit_predict(model, X)"
       ]
     },
     {
@@ -192,7 +192,7 @@
       },
       "outputs": [],
       "source": [
-        "y_true[\"cardiotocography\"] = y\nfor model_name in model_names:\n    model = make_estimator(\n        name=model_name,\n        lof_kw={\"n_neighbors\": int(n_samples * anomaly_frac)},\n        iforest_kw={\"random_state\": 42},\n    )\n    y_pred[model_name][\"cardiotocography\"] = fit_predict(model, X)"
+        "y_true[\"cardiotocography\"] = y\nfor model_name in model_names:\n    model = make_estimator(\n        name=model_name,\n        lof_kw={\"n_neighbors\": int(n_samples * anomaly_frac)},\n        iforest_kw={\"random_state\": 42},\n    )\n    y_score[model_name][\"cardiotocography\"] = fit_predict(model, X)"
       ]
     },
     {
@@ -210,7 +210,7 @@
       },
       "outputs": [],
       "source": [
-        "import math\n\nfrom sklearn.metrics import RocCurveDisplay\n\ncols = 2\npos_label = 0  # mean 0 belongs to positive class\ndatasets_names = y_true.keys()\nrows = math.ceil(len(datasets_names) / cols)\n\nfig, axs = plt.subplots(nrows=rows, ncols=cols, squeeze=False, figsize=(10, rows * 4))\n\nfor ax, dataset_name in zip(axs.ravel(), datasets_names):\n    for model_idx, model_name in enumerate(model_names):\n        display = RocCurveDisplay.from_predictions(\n            y_true[dataset_name],\n            y_pred[model_name][dataset_name],\n            pos_label=pos_label,\n            name=model_name,\n            ax=ax,\n            plot_chance_level=(model_idx == len(model_names) - 1),\n            chance_level_kw={\"linestyle\": \":\"},\n        )\n    ax.set_title(dataset_name)\n_ = plt.tight_layout(pad=2.0)  # spacing between subplots"
+        "import math\n\nfrom sklearn.metrics import RocCurveDisplay\n\ncols = 2\npos_label = 0  # mean 0 belongs to positive class\ndatasets_names = y_true.keys()\nrows = math.ceil(len(datasets_names) / cols)\n\nfig, axs = plt.subplots(nrows=rows, ncols=cols, squeeze=False, figsize=(10, rows * 4))\n\nfor ax, dataset_name in zip(axs.ravel(), datasets_names):\n    for model_idx, model_name in enumerate(model_names):\n        display = RocCurveDisplay.from_predictions(\n            y_true[dataset_name],\n            y_score[model_name][dataset_name],\n            pos_label=pos_label,\n            name=model_name,\n            ax=ax,\n            plot_chance_level=(model_idx == len(model_names) - 1),\n            chance_level_kw={\"linestyle\": \":\"},\n        )\n    ax.set_title(dataset_name)\n_ = plt.tight_layout(pad=2.0)  # spacing between subplots"
       ]
     },
     {
@@ -228,7 +228,7 @@
       },
       "outputs": [],
       "source": [
-        "X = X_forestcover\ny = y_true[\"forestcover\"]\n\nn_samples = X.shape[0]\nn_neighbors_list = (n_samples * np.array([0.2, 0.02, 0.01, 0.001])).astype(np.int32)\nmodel = make_pipeline(RobustScaler(), LocalOutlierFactor())\n\nlinestyles = [\"solid\", \"dashed\", \"dashdot\", \":\", (5, (10, 3))]\n\nfig, ax = plt.subplots()\nfor model_idx, (linestyle, n_neighbors) in enumerate(zip(linestyles, n_neighbors_list)):\n    model.set_params(localoutlierfactor__n_neighbors=n_neighbors)\n    model.fit(X)\n    y_pred = model[-1].negative_outlier_factor_\n    display = RocCurveDisplay.from_predictions(\n        y,\n        y_pred,\n        pos_label=pos_label,\n        name=f\"n_neighbors = {n_neighbors}\",\n        ax=ax,\n        plot_chance_level=(model_idx == len(n_neighbors_list) - 1),\n        chance_level_kw={\"linestyle\": (0, (1, 10))},\n        linestyle=linestyle,\n        linewidth=2,\n    )\n_ = ax.set_title(\"RobustScaler with varying n_neighbors\\non forestcover dataset\")"
+        "X = X_forestcover\ny = y_true[\"forestcover\"]\n\nn_samples = X.shape[0]\nn_neighbors_list = (n_samples * np.array([0.2, 0.02, 0.01, 0.001])).astype(np.int32)\nmodel = make_pipeline(RobustScaler(), LocalOutlierFactor())\n\nlinestyles = [\"solid\", \"dashed\", \"dashdot\", \":\", (5, (10, 3))]\n\nfig, ax = plt.subplots()\nfor model_idx, (linestyle, n_neighbors) in enumerate(zip(linestyles, n_neighbors_list)):\n    model.set_params(localoutlierfactor__n_neighbors=n_neighbors)\n    model.fit(X)\n    y_score = model[-1].negative_outlier_factor_\n    display = RocCurveDisplay.from_predictions(\n        y,\n        y_score,\n        pos_label=pos_label,\n        name=f\"n_neighbors = {n_neighbors}\",\n        ax=ax,\n        plot_chance_level=(model_idx == len(n_neighbors_list) - 1),\n        chance_level_kw={\"linestyle\": (0, (1, 10))},\n        linestyle=linestyle,\n        linewidth=2,\n    )\n_ = ax.set_title(\"RobustScaler with varying n_neighbors\\non forestcover dataset\")"
       ]
     },
     {
@@ -246,7 +246,7 @@
       },
       "outputs": [],
       "source": [
-        "from sklearn.preprocessing import MinMaxScaler, SplineTransformer, StandardScaler\n\npreprocessor_list = [\n    None,\n    RobustScaler(),\n    StandardScaler(),\n    MinMaxScaler(),\n    SplineTransformer(),\n]\nexpected_anomaly_fraction = 0.02\nlof = LocalOutlierFactor(n_neighbors=int(n_samples * expected_anomaly_fraction))\n\nfig, ax = plt.subplots()\nfor model_idx, (linestyle, preprocessor) in enumerate(\n    zip(linestyles, preprocessor_list)\n):\n    model = make_pipeline(preprocessor, lof)\n    model.fit(X)\n    y_pred = model[-1].negative_outlier_factor_\n    display = RocCurveDisplay.from_predictions(\n        y,\n        y_pred,\n        pos_label=pos_label,\n        name=str(preprocessor).split(\"(\")[0],\n        ax=ax,\n        plot_chance_level=(model_idx == len(preprocessor_list) - 1),\n        chance_level_kw={\"linestyle\": (0, (1, 10))},\n        linestyle=linestyle,\n        linewidth=2,\n    )\n_ = ax.set_title(\"Fixed n_neighbors with varying preprocessing\\non forestcover dataset\")"
+        "from sklearn.preprocessing import MinMaxScaler, SplineTransformer, StandardScaler\n\npreprocessor_list = [\n    None,\n    RobustScaler(),\n    StandardScaler(),\n    MinMaxScaler(),\n    SplineTransformer(),\n]\nexpected_anomaly_fraction = 0.02\nlof = LocalOutlierFactor(n_neighbors=int(n_samples * expected_anomaly_fraction))\n\nfig, ax = plt.subplots()\nfor model_idx, (linestyle, preprocessor) in enumerate(\n    zip(linestyles, preprocessor_list)\n):\n    model = make_pipeline(preprocessor, lof)\n    model.fit(X)\n    y_score = model[-1].negative_outlier_factor_\n    display = RocCurveDisplay.from_predictions(\n        y,\n        y_score,\n        pos_label=pos_label,\n        name=str(preprocessor).split(\"(\")[0],\n        ax=ax,\n        plot_chance_level=(model_idx == len(preprocessor_list) - 1),\n        chance_level_kw={\"linestyle\": (0, (1, 10))},\n        linestyle=linestyle,\n        linewidth=2,\n    )\n_ = ax.set_title(\"Fixed n_neighbors with varying preprocessing\\non forestcover dataset\")"
       ]
     },
     {
@@ -264,7 +264,7 @@
       },
       "outputs": [],
       "source": [
-        "X = X_cardiotocography\ny = y_true[\"cardiotocography\"]\n\nn_samples, expected_anomaly_fraction = X.shape[0], 0.025\nlof = LocalOutlierFactor(n_neighbors=int(n_samples * expected_anomaly_fraction))\n\nfig, ax = plt.subplots()\nfor model_idx, (linestyle, preprocessor) in enumerate(\n    zip(linestyles, preprocessor_list)\n):\n    model = make_pipeline(preprocessor, lof)\n    model.fit(X)\n    y_pred = model[-1].negative_outlier_factor_\n    display = RocCurveDisplay.from_predictions(\n        y,\n        y_pred,\n        pos_label=pos_label,\n        name=str(preprocessor).split(\"(\")[0],\n        ax=ax,\n        plot_chance_level=(model_idx == len(preprocessor_list) - 1),\n        chance_level_kw={\"linestyle\": (0, (1, 10))},\n        linestyle=linestyle,\n        linewidth=2,\n    )\nax.set_title(\n    \"Fixed n_neighbors with varying preprocessing\\non cardiotocography dataset\"\n)\nplt.show()"
+        "X = X_cardiotocography\ny = y_true[\"cardiotocography\"]\n\nn_samples, expected_anomaly_fraction = X.shape[0], 0.025\nlof = LocalOutlierFactor(n_neighbors=int(n_samples * expected_anomaly_fraction))\n\nfig, ax = plt.subplots()\nfor model_idx, (linestyle, preprocessor) in enumerate(\n    zip(linestyles, preprocessor_list)\n):\n    model = make_pipeline(preprocessor, lof)\n    model.fit(X)\n    y_score = model[-1].negative_outlier_factor_\n    display = RocCurveDisplay.from_predictions(\n        y,\n        y_score,\n        pos_label=pos_label,\n        name=str(preprocessor).split(\"(\")[0],\n        ax=ax,\n        plot_chance_level=(model_idx == len(preprocessor_list) - 1),\n        chance_level_kw={\"linestyle\": (0, (1, 10))},\n        linestyle=linestyle,\n        linewidth=2,\n    )\nax.set_title(\n    \"Fixed n_neighbors with varying preprocessing\\non cardiotocography dataset\"\n)\nplt.show()"
       ]
     }
   ],