Testing frequency

hands_on/pyanno_voting/pyanno/tests/test_voting.py

@@ -2,9 +2,11 @@
 
 from pyanno import voting
 from pyanno.voting import MISSING_VALUE as MV
+from numpy.testing import assert_array_almost_equal
 
 
 def test_labels_count():
+    # Given
     annotations = [
         [1,  2, MV, MV],
         [MV, MV,  3,  3],
@@ -13,7 +15,11 @@ def test_labels_count():
     ]
     nclasses = 5
     expected = [0, 3, 1, 3, 0]
+
+    # When
     result = voting.labels_count(annotations, nclasses)
+
+    # Then
     assert result == expected
 
 
@@ -41,3 +47,16 @@ def test_majority_vote_empty_item():
     expected = [1, MV, 2]
     result = voting.majority_vote(annotations)
     assert result == expected
+
+def test_label_frequency():
+    # Given 
+    matrix = [[1, 1, 2], [-1, 1, 2]]
+    classes = 4
+    expected_result = np.array([ 0. ,  0.6,  0.4,  0. ])
+
+    # When
+    function_result = voting.labels_frequency([[1, 1, 2], [-1, 1, 2]], 4)
+
+    # Then
+    assert_array_almost_equal(function_result, expected_result, decimal=6)
+

hands_on/pyanno_voting/pyanno/voting.py

@@ -100,3 +100,33 @@ def labels_frequency(annotations, nclasses):
         freq[k] is the frequency of elements of class k in `annotations`, i.e.
         their count over the number of total of observed (non-missing) elements
     """
+    # Transform list of lists into a np array
+    annot = np.array(annotations)
+
+    # Compute the frequency of each value in the matrix 
+    unique, counts = np.unique(annot, return_counts=True)
+
+    # Create a dictionary with the frequency values 
+    freq = dict(zip(unique, counts))
+
+    # Good values 
+    good_values = annot.size - freq[-1]
+
+
+    # Remove -1 freq from the dict
+    freq.pop(-1, None)
+
+
+    freq_list = []
+    classes = list(range(nclasses))
+
+    for class_value in classes: 
+        if class_value in freq.keys():
+            freq_list.append(freq[class_value]/good_values)
+
+        else: 
+            freq_list.append(0)
+
+    return np.array(freq_list)
+
+