Fixes in ENN

Author: jhosoume

.Rhistory

@@ -5,16 +5,16 @@
 from NoiseFiltersPy.Filter import *
 
 class ENN:
-    def __init__(self, max_neighbours = ):
-        self.max_neighbours = max_neighbours
-        self.filter = Filter(parameters = {"max_neighbours": self.max_neighbours})
+    def __init__(self, neighbours = 3):
+        self.neighbours = neighbours
+        self.filter = Filter(parameters = {"neighbours": self.neighbours})
 
     def __call__(self, data, classes):
         self.isNoise = np.array([False] * len(classes))
-        self.clf = KNeighborsClassifier(n_neighbors = n_neigh, algorithm = 'kd_tree', n_jobs = -1)
+        self.clf = KNeighborsClassifier(n_neighbors = self.neighbours, algorithm = 'kd_tree', n_jobs = -1)
         for indx in range(len(data)):
             self.clf.fit(np.delete(data, indx, axis = 0), np.delete(classes, indx, axis = 0))
-            pred = self.clf.predict(data[indx])
+            pred = self.clf.predict(data[indx].reshape(1, -1))
             self.isNoise[indx] = pred != classes[indx]
         self.filter.remIndx = np.argwhere(self.isNoise)
         notNoise = np.invert(self.isNoise)

NoiseFiltersPy/ENN.py

@@ -0,0 +1,8 @@
+from sklearn import datasets
+from NoiseFiltersPy.ENN import ENN
+
+dataset = datasets.load_iris()
+data = dataset.data
+classes = dataset.target
+enn = ENN()
+filter = enn(data, classes)

examples/enn_example.py

@@ -5,7 +5,7 @@
 
 setuptools.setup(
     name="NoiseFiltersPy",
-    version="0.0.4",
+    version="0.0.5",
     author="Juliana Hosoume and Luis Faina",
     author_email="[email protected]",
     description="Python implementation of NoiseFiltersR",