Merge pull request #247 from yzhao062/development

V0.8.4

CHANGES.txt

@@ -108,6 +108,10 @@ v<0.8.2>, <07/04/2020> -- Add a set of utility functions.
 v<0.8.2>, <08/30/2020> -- Add COPOD and MAD algorithm.
 v<0.8.3>, <09/01/2020> -- Make decision score consistent.
 v<0.8.3>, <09/19/2020> -- Add model persistence documentation (save and load).
+v<0.8.4>, <10/13/2020> -- Fix COPOD code inconsistency (issue #239).
+v<0.8.4>, <10/24/2020> -- Fix LSCP minor bug (issue #180).
+v<0.8.4>, <11/02/2020> -- Add support for Tensorflow 2.
+v<0.8.4>, <11/12/2020> -- Merge PR #!02 for categortical data generation.
 
 
 

examples/generate_data_categorical_example.py

@@ -0,0 +1,49 @@
+# -*- coding: utf-8 -*-
+"""Example of using and visualizing ``generate_data_categorical`` function.
+"""
+# Author: Yahya Almardeny <almardeny@gmail.com>
+# License: BSD 2 clause
+
+from __future__ import division
+from __future__ import print_function
+
+import os
+import sys
+import numpy as np
+import matplotlib.pyplot as plt
+
+# temporary solution for relative imports in case pyod is not installed
+# if pyod is installed, no need to use the following line
+
+sys.path.append(
+    os.path.abspath(os.path.join(os.path.dirname("__file__"), '..')))
+
+from pyod.utils.data import generate_data_categorical
+
+if __name__ == "__main__":
+    contamination = 0.1  # percentage of outliers
+
+    # Generate sample data in clusters
+    X_train, X_test, y_train, y_test = generate_data_categorical \
+        (n_train=200, n_test=50,
+         n_category_in=8, n_category_out=5,
+         n_informative=1, n_features=1,
+         contamination=contamination,
+         shuffle=True, random_state=42)
+
+    # note that visalizing it can only be in 1 dimension!
+    cats = list(np.ravel(X_train))
+    labels = list(y_train)
+    fig, axs = plt.subplots(1, 2)
+    axs[0].bar(cats, labels)
+    axs[1].plot(cats, labels)
+    plt.title('Synthetic Categorical Train Data')
+    plt.show()
+
+    cats = list(np.ravel(X_test))
+    labels = list(y_test)
+    fig, axs = plt.subplots(1, 2)
+    axs[0].bar(cats, labels)
+    axs[1].plot(cats, labels)
+    plt.title('Synthetic Categorical Test Data')
+    plt.show()

pyod/models/auto_encoder.py

@@ -8,10 +8,6 @@
 from __future__ import print_function
 
 import numpy as np
-from keras.models import Sequential
-from keras.layers import Dense, Dropout
-from keras.regularizers import l2
-from keras.losses import mean_squared_error
 from sklearn.preprocessing import StandardScaler
 from sklearn.utils import check_array
 from sklearn.utils.validation import check_is_fitted
@@ -20,6 +16,19 @@
 from ..utils.stat_models import pairwise_distances_no_broadcast
 
 from .base import BaseDetector
+from .base_dl import _get_tensorflow_version
+
+# if tensorflow 2, import from tf directly
+if _get_tensorflow_version() == 1:
+    from keras.models import Sequential
+    from keras.layers import Dense, Dropout
+    from keras.regularizers import l2
+    from keras.losses import mean_squared_error
+else:
+    from tensorflow.keras.models import Sequential
+    from tensorflow.keras.layers import Dense, Dropout
+    from tensorflow.keras.regularizers import l2
+    from tensorflow.keras.losses import mean_squared_error
 
 
 # noinspection PyUnresolvedReferences,PyPep8Naming,PyTypeChecker
@@ -78,7 +87,7 @@ class AutoEncoder(BaseDetector):
         - 1 = progress bar
         - 2 = one line per epoch.
 
-        For verbosity >= 1, model summary may be printed.
+        For verbose >= 1, model summary may be printed.
 
     random_state : random_state: int, RandomState instance or None, optional
         (default=None)

pyod/models/base_dl.py

@@ -0,0 +1,27 @@
+# -*- coding: utf-8 -*-
+"""Base class for deep learning models
+"""
+# Author: Yue Zhao <zhaoy@cmu.edu>
+# License: BSD 2 clause
+
+from __future__ import division
+from __future__ import print_function
+
+import tensorflow
+
+def _get_tensorflow_version():  # pragma: no cover
+    """ Utility function to decide the version of tensorflow, which will 
+    affect how to import keras models. 
+
+    Returns
+    -------
+    tensorflow version : int
+
+    """
+
+    tf_version = str(tensorflow.__version__)
+    if int(tf_version.split(".")[0]) != 1 and int(
+            tf_version.split(".")[0]) != 2:
+        raise ValueError("tensorflow version error")
+
+    return int(tf_version.split(".")[0])

pyod/models/copod.py

@@ -79,9 +79,10 @@ def fit(self, X, y=None):
             Fitted estimator.
         """
         X = check_array(X)
-        self._set_n_classes(y=None)
+        self._set_n_classes(y)
         self.X_train = X
         self.decision_function(X)
+        return self
 
     def decision_function(self, X):
         """Predict raw anomaly score of X using the fitted detector.

pyod/models/gaal_base.py

@@ -11,9 +11,17 @@
 
 import math
 
-import keras
-from keras.layers import Input, Dense
-from keras.models import Sequential, Model
+from .base_dl import _get_tensorflow_version
+
+# if tensorflow 2, import from tf directly
+if _get_tensorflow_version() == 1:
+    import keras
+    from keras.layers import Input, Dense
+    from keras.models import Sequential, Model
+else:
+    import tensorflow.keras as keras
+    from tensorflow.keras.layers import Input, Dense
+    from tensorflow.keras.models import Sequential, Model
 
 
 # TODO: create a base class for so_gaal and mo_gaal

pyod/models/lscp.py

@@ -341,10 +341,18 @@ def _get_local_region(self, X_test_norm):
         # keep nearby points which occur at least local_region_threshold times
         final_local_region_list = [[]] * X_test_norm.shape[0]
         for j in range(X_test_norm.shape[0]):
-            final_local_region_list[j] = [item for item, count in
-                                          collections.Counter(
-                                              local_region_list[j]).items() if
-                                          count > self.local_region_threshold]
+            tmp = [item for item, count in collections.Counter(
+                local_region_list[j]).items() if
+                   count > self.local_region_threshold]
+            decrease_value = 0
+            while len(tmp) < 2:
+                decrease_value = decrease_value + 1
+                assert decrease_value < self.local_region_threshold
+                tmp = [item for item, count in
+                       collections.Counter(local_region_list[j]).items() if
+                       count > (self.local_region_threshold - decrease_value)]
+
+            final_local_region_list[j] = tmp
 
         return final_local_region_list
 

pyod/models/mo_gaal.py

@@ -13,16 +13,23 @@
 
 import numpy as np
 
-from keras.layers import Input
-from keras.models import Model
-from keras.optimizers import SGD
-
 from sklearn.utils import check_array
 from sklearn.utils.validation import check_is_fitted
 
 from .base import BaseDetector
 from .gaal_base import create_discriminator
 from .gaal_base import create_generator
+from .base_dl import _get_tensorflow_version
+
+# if tensorflow 2, import from tf directly
+if _get_tensorflow_version() == 1:
+    from keras.layers import Input
+    from keras.models import Model
+    from keras.optimizers import SGD
+else:
+    from tensorflow.keras.layers import Input
+    from tensorflow.keras.models import Model
+    from tensorflow.keras.optimizers import SGD
 
 
 class MO_GAAL(BaseDetector):

pyod/models/so_gaal.py

@@ -13,16 +13,23 @@
 
 import numpy as np
 
-from keras.layers import Input
-from keras.models import Model
-from keras.optimizers import SGD
-
 from sklearn.utils import check_array
 from sklearn.utils.validation import check_is_fitted
 
 from .base import BaseDetector
 from .gaal_base import create_discriminator
 from .gaal_base import create_generator
+from .base_dl import _get_tensorflow_version
+
+# if tensorflow 2, import from tf directly
+if _get_tensorflow_version() == 1:
+    from keras.layers import Input
+    from keras.models import Model
+    from keras.optimizers import SGD
+else:
+    from tensorflow.keras.layers import Input
+    from tensorflow.keras.models import Model
+    from tensorflow.keras.optimizers import SGD
 
 
 class SO_GAAL(BaseDetector):

pyod/models/vae.py

@@ -22,13 +22,6 @@
 
 import numpy as np
 
-from keras.models import Model
-from keras.layers import Lambda, Input, Dense, Dropout
-from keras.regularizers import l2
-from keras.losses import mse, binary_crossentropy
-from keras.utils import plot_model
-from keras import backend as K
-
 from sklearn.preprocessing import StandardScaler
 from sklearn.utils import check_array
 from sklearn.utils.validation import check_is_fitted
@@ -37,6 +30,21 @@
 from ..utils.stat_models import pairwise_distances_no_broadcast
 
 from .base import BaseDetector
+from .base_dl import _get_tensorflow_version
+
+# if tensorflow 2, import from tf directly
+if _get_tensorflow_version() == 1:
+    from keras.models import Model
+    from keras.layers import Lambda, Input, Dense, Dropout
+    from keras.regularizers import l2
+    from keras.losses import mse, binary_crossentropy
+    from keras import backend as K
+else:
+    from tensorflow.keras.models import Model
+    from tensorflow.keras.layers import Lambda, Input, Dense, Dropout
+    from tensorflow.keras.regularizers import l2
+    from tensorflow.keras.losses import mse, binary_crossentropy
+    from tensorflow.keras import backend as K
 
 
 class VAE(BaseDetector):
@@ -114,13 +122,13 @@ class VAE(BaseDetector):
         If True, apply standardization on the data.
 
     verbose : int, optional (default=1)
-        Verbosity mode.
+        verbose mode.
 
         - 0 = silent
         - 1 = progress bar
         - 2 = one line per epoch.
 
-        For verbosity >= 1, model summary may be printed.
+        For verbose >= 1, model summary may be printed.
 
     random_state : random_state: int, RandomState instance or None, opti
         (default=None)
@@ -172,7 +180,7 @@ def __init__(self, encoder_neurons=None, decoder_neurons=None,
                  output_activation='sigmoid', loss=mse, optimizer='adam',
                  epochs=100, batch_size=32, dropout_rate=0.2,
                  l2_regularizer=0.1, validation_size=0.1, preprocessing=True,
-                 verbosity=1, random_state=None, contamination=0.1,
+                 verbose=1, random_state=None, contamination=0.1,
                  gamma=1.0, capacity=0.0):
         super(VAE, self).__init__(contamination=contamination)
         self.encoder_neurons = encoder_neurons
@@ -187,7 +195,7 @@ def __init__(self, encoder_neurons=None, decoder_neurons=None,
         self.l2_regularizer = l2_regularizer
         self.validation_size = validation_size
         self.preprocessing = preprocessing
-        self.verbosity = verbosity
+        self.verbose = verbose
         self.random_state = random_state
         self.latent_dim = latent_dim
         self.gamma = gamma
@@ -264,7 +272,7 @@ def _build_model(self):
             [z_mean, z_log])
         # Instantiate encoder
         encoder = Model(inputs, [z_mean, z_log, z])
-        if self.verbosity >= 1:
+        if self.verbose >= 1:
             encoder.summary()
 
         # Build Decoder
@@ -281,7 +289,7 @@ def _build_model(self):
             layer)
         # Instatiate decoder
         decoder = Model(latent_inputs, outputs)
-        if self.verbosity >= 1:
+        if self.verbose >= 1:
             decoder.summary()
         # Generate outputs
         outputs = decoder(encoder(inputs)[2])
@@ -290,7 +298,7 @@ def _build_model(self):
         vae = Model(inputs, outputs)
         vae.add_loss(self.vae_loss(inputs, outputs, z_mean, z_log))
         vae.compile(optimizer=self.optimizer)
-        if self.verbosity >= 1:
+        if self.verbose >= 1:
             vae.summary()
         return vae
 
@@ -335,7 +343,7 @@ def fit(self, X, y=None):
                                         batch_size=self.batch_size,
                                         shuffle=True,
                                         validation_split=self.validation_size,
-                                        verbose=self.verbosity).history
+                                        verbose=self.verbose).history
         # Predict on X itself and calculate the reconstruction error as
         # the outlier scores. Noted X_norm was shuffled has to recreate
         if self.preprocessing: