remove main function in each model

.gitignore

@@ -2,5 +2,6 @@ deepod.egg-info/
 dist/
 build/
 .idea
-#data/
+deepod/models/_local_test_.py
+data/
 **/__pycache__

README.rst

@@ -17,16 +17,14 @@ Python Deep Outlier/Anomaly Detection (DeepOD)
 **DeepOD** is an open-source python framework for deep learning-based anomaly detection on multivariate data. DeepOD provides unified low-code implementation of different detection models based on PyTorch.
 
 
-DeepOD includes nine popular deep outlier detection / anomaly detection algorithms (in unsupervised/weakly-supervised paradigm) for now. More baseline algorithms will be included later.
+DeepOD includes ten popular deep outlier detection / anomaly detection algorithms (in unsupervised/weakly-supervised paradigm) for now. More baseline algorithms will be included later.
 
 
 
 Installation
 ~~~~~~~~~~~~~~
 The DeepOD framework can be installed via:
 
-install a stable version (we have six models in this version, please clone this repository for more models)
-
 
 .. code-block:: bash
 
@@ -64,7 +62,6 @@ Supported Models
  DevNet, KDD, 2019, weakly-supervised, Deep Anomaly Detection with Deviation Networks
  PReNet, ArXiv, 2020, weakly-supervised, Deep Weakly-supervised Anomaly Detection
  Deep SAD, ICLR, 2020, weakly-supervised, Deep Semi-Supervised Anomaly Detection
-~~~~~~~~~~~~~~
 
 
 Usages

deepod/models/_local_test_.py

@@ -0,0 +1,32 @@
+from deepod.models import *
+from sklearn.metrics import roc_auc_score
+import numpy as np
+
+if __name__ == '__main__':
+    import pandas as pd
+    import numpy as np
+
+    file = '../../data/38_thyroid.npz'
+    data = np.load(file, allow_pickle=True)
+    x, y = data['X'], data['y']
+    y = np.array(y, dtype=int)
+
+    anom_id = np.where(y==1)[0]
+    known_anom_id = np.random.choice(anom_id, 30)
+    y_semi = np.zeros_like(y)
+    y_semi[known_anom_id] = 1
+
+    clf = DevNet(device='cpu')
+    clf.fit(x, y_semi)
+
+    scores = clf.decision_function(x)
+
+    from sklearn.metrics import roc_auc_score
+    auc = roc_auc_score(y_score=scores, y_true=y)
+    print(auc)
+
+    clf = DeepSVDD(device='cpu')
+    clf.fit(x, y_semi)
+    scores = clf.decision_function(x)
+    auc = roc_auc_score(y_score=scores, y_true=y)
+    print(auc)

deepod/models/anogan.py

@@ -209,27 +209,3 @@ def training_forward(self, batch_x, net, criterion):
     def inference_forward(self, batch_x, net, criterion):
         # implement in _inference
         pass
-
-if __name__ == '__main__':
-    import numpy as np
-
-    file = '../../data/38_thyroid.npz'
-    data = np.load(file, allow_pickle=True)
-    x, y = data['X'], data['y']
-    y = np.array(y, dtype=int)
-
-    anom_id = np.where(y==1)[0]
-    known_anom_id = np.random.choice(anom_id, 30)
-    y_semi = np.zeros_like(y)
-    y_semi[known_anom_id] = 1
-
-    clf = AnoGAN(device='cuda', prt_steps=1, epochs=100)
-    clf.fit(x, y_semi)
-
-    scores = clf.decision_function(x)
-
-    from sklearn.metrics import roc_auc_score
-
-    auc = roc_auc_score(y_score=scores, y_true=y)
-
-    print(auc)

deepod/models/devnet.py

@@ -172,29 +172,3 @@ def forward(self, y_true, y_pred):
             return loss
 
         return loss
-
-
-if __name__ == '__main__':
-    import pandas as pd
-    import numpy as np
-
-    file = '../../data/38_thyroid.npz'
-    data = np.load(file, allow_pickle=True)
-    x, y = data['X'], data['y']
-    y = np.array(y, dtype=int)
-
-    anom_id = np.where(y==1)[0]
-    known_anom_id = np.random.choice(anom_id, 30)
-    y_semi = np.zeros_like(y)
-    y_semi[known_anom_id] = 1
-
-    clf = DevNet(device='cpu')
-    clf.fit(x, y_semi)
-
-    scores = clf.decision_function(x)
-
-    from sklearn.metrics import roc_auc_score
-
-    auc = roc_auc_score(y_score=scores, y_true=y)
-
-    print(auc)

deepod/models/dsad.py

@@ -187,28 +187,3 @@ def forward(self, rep, semi_targets=None, reduction=None):
             return torch.sum(loss)
         elif reduction == 'none':
             return loss
-
-
-if __name__ == '__main__':
-    import numpy as np
-
-    file = '../../data/38_thyroid.npz'
-    data = np.load(file, allow_pickle=True)
-    x, y = data['X'], data['y']
-    y = np.array(y, dtype=int)
-
-    anom_id = np.where(y == 1)[0]
-    known_anom_id = np.random.choice(anom_id, 30)
-    y_semi = np.zeros_like(y)
-    y_semi[known_anom_id] = 1
-
-    clf = DeepSAD(device='cpu')
-    clf.fit(x, y_semi)
-
-    scores = clf.decision_function(x)
-
-    from sklearn.metrics import roc_auc_score
-
-    auc = roc_auc_score(y_score=scores, y_true=y)
-
-    print(auc)

deepod/models/dsvdd.py

@@ -166,28 +166,3 @@ def forward(self, rep, reduction=None):
             return torch.sum(loss)
         elif reduction == 'none':
             return loss
-
-
-if __name__ == '__main__':
-    import numpy as np
-
-    file = '../../data/38_thyroid.npz'
-    data = np.load(file, allow_pickle=True)
-    x, y = data['X'], data['y']
-    y = np.array(y, dtype=int)
-
-    anom_id = np.where(y==1)[0]
-    known_anom_id = np.random.choice(anom_id, 30)
-    y_semi = np.zeros_like(y)
-    y_semi[known_anom_id] = 1
-
-    clf = DeepSVDD(device='cpu')
-    clf.fit(x, y_semi)
-
-    scores = clf.decision_function(x)
-
-    from sklearn.metrics import roc_auc_score
-
-    auc = roc_auc_score(y_score=scores, y_true=y)
-
-    print(auc)

deepod/models/goad.py

@@ -192,31 +192,3 @@ def weights_init(m):
         torch.nn.init.eye_(m.weight)
     elif classname.find('Emb') != -1:
         torch.nn.init.normal(m.weight, mean=0, std=0.01)
-
-
-
-
-if __name__ == '__main__':
-    import pandas as pd
-    import numpy as np
-
-    file = '../../data/38_thyroid.npz'
-    data = np.load(file, allow_pickle=True)
-    x, y = data['X'], data['y']
-    y = np.array(y, dtype=int)
-
-    anom_id = np.where(y==1)[0]
-    known_anom_id = np.random.choice(anom_id, 30)
-    y_semi = np.zeros_like(y)
-    y_semi[known_anom_id] = 1
-
-    clf = GOAD(device='cpu', epochs=1)
-    clf.fit(x, y_semi)
-
-    scores = clf.decision_function(x)
-
-    from sklearn.metrics import roc_auc_score
-
-    auc = roc_auc_score(y_score=scores, y_true=y)
-
-    print(auc)

deepod/models/icl.py

@@ -276,47 +276,3 @@ def positive_matrix_builder(self, data):
         complement_matrix = data[:, np.arange(self.all_idx.shape[0])[:, None], self.all_idx_complement]
 
         return matrix, complement_matrix
-
-
-if __name__ == '__main__':
-    import numpy as np
-    import pandas as pd
-
-    file = '/home/xuhz/dataset/1-tabular/fmnist.inlier_var.3percent.5dup/fmnist_anom7_nnormal9_id9_1.csv'
-    df = pd.read_csv(file)
-    df.replace([np.inf, -np.inf], np.nan, inplace=True)
-    df.fillna(method='ffill', inplace=True)
-    x = df.values[:, :-1]
-    y = np.array(df.values[:, -1], dtype=int)
-
-    norm_idx = np.where(y == 0)[0]
-    anom_idx = np.where(y == 1)[0]
-    split = int(0.5 * len(norm_idx))
-    train_norm_idx, test_norm_idx = norm_idx[:split], norm_idx[split:]
-
-    x_train = x[train_norm_idx]
-    y_train = y[train_norm_idx]
-
-    x_test = x[np.hstack([test_norm_idx, anom_idx])]
-    y_test = y[np.hstack([test_norm_idx, anom_idx])]
-    x_train = x_train / 255
-    x_test = x_test / 255
-
-    # file = '../../data/38_thyroid.npz'
-    # data_ = np.load(file, allow_pickle=True)
-    # x, y = data_['X'], data_['y']
-    # y = np.array(y, dtype=int)
-    # x_train = x
-    # x_test = x
-    # y_test = y
-
-    clf = ICL(device='cuda', epochs=100, hidden_dims='100', act='LeakyReLU', verbose=2)
-    clf.fit(x_train)
-
-    scores = clf.decision_function(x_test)
-
-    from sklearn.metrics import roc_auc_score
-
-    auc = roc_auc_score(y_score=scores, y_true=y_test)
-
-    print(auc)

deepod/models/neutral.py

@@ -191,30 +191,3 @@ def forward(self, z):
             return loss
 
         return loss
-
-
-
-if __name__ == '__main__':
-    import pandas as pd
-    import numpy as np
-
-    file = '../../data/38_thyroid.npz'
-    data = np.load(file, allow_pickle=True)
-    x, y = data['X'], data['y']
-    y = np.array(y, dtype=int)
-
-    anom_id = np.where(y==1)[0]
-    known_anom_id = np.random.choice(anom_id, 30)
-    y_semi = np.zeros_like(y)
-    y_semi[known_anom_id] = 1
-
-    clf = NeuTraL(device='cpu')
-    clf.fit(x, y_semi)
-
-    scores = clf.decision_function(x)
-
-    from sklearn.metrics import roc_auc_score
-
-    auc = roc_auc_score(y_score=scores, y_true=y)
-
-    print(auc)

deepod/models/prenet.py

@@ -201,30 +201,3 @@ def batch_generation(self):
                 batch_y[i] = 8
 
         return batch_x1, batch_x2, batch_y
-
-
-if __name__ == '__main__':
-    import pandas as pd
-    import numpy as np
-
-    # file = '../../data/38_thyroid.npz'
-    file = '/home/xuhz/dataset/1-tabular/ADBench-classical/27_PageBlocks.npz'
-    data = np.load(file, allow_pickle=True)
-    x, y = data['X'], data['y']
-    y = np.array(y, dtype=int)
-
-    anom_id = np.where(y == 1)[0]
-    k_anom_id = np.random.choice(anom_id, 30, replace=False)
-    y_semi = np.zeros_like(y)
-    y_semi[k_anom_id] = 1
-
-    clf = PReNet(device='cpu', batch_size=256, epochs=10)
-    clf.fit(x, y_semi)
-
-    scores = clf.decision_function(x)
-
-    from sklearn.metrics import roc_auc_score
-
-    auc = roc_auc_score(y_score=scores, y_true=y)
-
-    print(auc)

deepod/models/rca.py

@@ -249,29 +249,3 @@ def forward(self, x1, x2):
         z2 = self.enc2(x2)
         x_hat2 = self.dec2(z2)
         return z1, z2, x_hat1, x_hat2
-
-
-if __name__ == '__main__':
-    import pandas as pd
-    import numpy as np
-
-    file = '../../data/38_thyroid.npz'
-    data = np.load(file, allow_pickle=True)
-    x, y = data['X'], data['y']
-    y = np.array(y, dtype=int)
-
-    anom_id = np.where(y==1)[0]
-    known_anom_id = np.random.choice(anom_id, 30)
-    y_semi = np.zeros_like(y)
-    y_semi[known_anom_id] = 1
-
-    clf = RCA(epochs=10, device='cpu')
-    clf.fit(x, y_semi)
-
-    scores = clf.decision_function(x)
-
-    from sklearn.metrics import roc_auc_score
-
-    auc = roc_auc_score(y_score=scores, y_true=y)
-
-    print(auc)

deepod/models/rdp.py

@@ -138,30 +138,3 @@ def forward(self, rep, rep1, x, x1):
             rdp_loss = F.mse_loss(d_target, d_pred, reduction='none')
 
         return gap_loss + rdp_loss
-
-
-
-if __name__ == '__main__':
-    import pandas as pd
-    import numpy as np
-
-    file = '../../data/38_thyroid.npz'
-    data = np.load(file, allow_pickle=True)
-    x, y = data['X'], data['y']
-    y = np.array(y, dtype=int)
-
-    anom_id = np.where(y==1)[0]
-    known_anom_id = np.random.choice(anom_id, 30)
-    y_semi = np.zeros_like(y)
-    y_semi[known_anom_id] = 1
-
-    clf = RDP(epochs=10, device='cpu')
-    clf.fit(x, y_semi)
-
-    scores = clf.decision_function(x)
-
-    from sklearn.metrics import roc_auc_score
-
-    auc = roc_auc_score(y_score=scores, y_true=y)
-
-    print(auc)