Remove duplicate points in principal_direction_extents (#999)

neurom/morphmath.py

@@ -441,10 +441,10 @@ def segment_taper_rate(seg):
 def pca(points):
     """Estimate the principal components of the covariance on the given point cloud.
 
-    Input
-        A numpy array of points of the form ((x1,y1,z1), (x2, y2, z2)...)
+    Args:
+        points: A numpy array of points of the form ((x1,y1,z1), (x2, y2, z2)...)
 
-    Ouptut
+    Returns:
         Eigenvalues and respective eigenvectors
     """
     return np.linalg.eig(np.cov(points.transpose()))
@@ -474,8 +474,10 @@ def principal_direction_extent(points):
         eigs : eigenvalues of the covariance matrix
         eigv : respective eigenvectors of the covariance matrix
     """
+    # pca can be biased by duplicate points
+    points = np.unique(points, axis=0)
+
     # center the points around 0.0
-    points = np.copy(points)
     points -= np.mean(points, axis=0)
 
     # principal components

tests/features/test_get_features.py

@@ -797,9 +797,15 @@ def test_principal_direction_extents():
 
     # test with a realistic morphology
     m = nm.load_morphology(DATA_PATH / 'h5/v1' / 'bio_neuron-000.h5')
-    p_ref = [1672.9694359427331, 142.43704397865031, 226.45895382204986,
-             415.50612748523838, 429.83008974193206, 165.95410536922873,
-             346.83281498399697]
+    p_ref = [
+        1672.969491,
+        142.437047,
+        224.607978,
+        415.50613,
+        429.830081,
+        165.954097,
+        346.832825,
+    ]
     p = features.get('principal_direction_extents', m)
     assert_allclose(p, p_ref, rtol=1e-6)