added DIF solver + SMAP dataset

datasets/glass.py

@@ -12,4 +12,4 @@ class Dataset(BaseDataset):
     def get_data(self):
         data = np.load("datasets/14_glass.npz")
         X, y = data["X"], data["y"]
-        return dict(X=X, y=y)
+        return dict(X=X, y=y, X_test=None)

datasets/simulated.py

@@ -1,7 +1,6 @@
 from benchopt import BaseDataset, safe_import_context
 
 with safe_import_context() as import_ctx:
-    # import module to generate normal 1d data
     from sklearn.datasets import make_regression
     import numpy as np
 

datasets/smap.py

@@ -0,0 +1,24 @@
+from benchopt import BaseDataset
+from benchopt import safe_import_context
+
+with safe_import_context() as import_ctx:
+    import os
+    import pandas as pd
+
+
+class Dataset(BaseDataset):
+    name = "SMAP"
+
+    install_cmd = "conda"
+    requirements = ["pandas"]
+
+    def get_data(self):
+
+        path = "/storage/store/work/jyehya/Benchmarks/processing/processed/SMAP"
+        dataset = "SMAP"
+
+        X_train = pd.read_pickle(os.path.join(path, dataset + "_train.pkl"))
+        X_test = pd.read_pickle(os.path.join(path, dataset + "_test.pkl"))
+        y_test = pd.read_pickle(os.path.join(path, dataset + "_test_label.pkl"))
+
+        return dict(X=X_train, y=y_test, X_test=X_test)

objective.py

@@ -14,9 +14,10 @@ def get_one_result(self):
         """
         return np.zeros(self.X.shape[0])
 
-    def set_data(self, X, y):
+    def set_data(self, X, y, X_test):
         "Set the data to compute the objective."
         self.X, self.y = X, y
+        self.X_test = X_test
 
     def evaluate_result(self, y_hat):
         "Evaluate the result provided by the solver."

solvers/dif.py

@@ -0,0 +1,33 @@
+# Deep Isolation Forest
+from benchopt import BaseSolver
+from benchopt import safe_import_context
+
+with safe_import_context() as import_ctx:
+    from pyod.models.dif import DIF
+
+
+class Solver(BaseSolver):
+    name = "DIF"
+
+    install_cmd = "conda"
+    requirements = ["pyod"]
+
+    parameters = {
+        "contamination": [0.05, 0.1, 0.2],
+    }
+
+    sampling_strategy = "run_once"
+
+    def set_objective(self, X, y, X_test=None):
+        # y is y_test, the learning is unsupervised
+        self.X = X
+        self.X_test = X_test
+        self.y = y
+
+    def run(self, _):
+        clf = DIF(contamination=self.contamination)
+        clf.fit(self.X)
+        self.y_hat = clf.predict(self.X_test)
+
+    def get_result(self):
+        return {"y_hat": self.y_hat}