Fixed #308 Transpose mstump/mstumped output

Author: seanlaw

docs/Tutorial_Multidimensional_Motif_Discovery.ipynb

@@ -192,9 +192,9 @@ def _discretize(a, bins, right=True):
     return np.digitize(a, bins, right=right)
 
 
-def _get_subspace(T, m, motif_idx, nn_idx, k, include=None, discords=False):
+def _get_subspace(T, m, subseq_idx, nn_idx, k, include=None, discords=False):
     """
-    Compute the multi-dimensional matrix profile subspace for a given motif index and
+    Compute the multi-dimensional matrix profile subspace for a given subseq index and
     its nearest neighbor index
 
     Parameters
@@ -206,8 +206,8 @@ def _get_subspace(T, m, motif_idx, nn_idx, k, include=None, discords=False):
     m : int
         Window size
 
-    motif_idx : int
-        The motif index in T
+    subseq_idx : int
+        The subsequence index in T
 
     nn_idx : int
         The nearest neighbor index in T
@@ -236,9 +236,9 @@ def _get_subspace(T, m, motif_idx, nn_idx, k, include=None, discords=False):
     """
     T, _, _ = core.preprocess(T, m)
 
-    motifs = core.z_norm(T[:, motif_idx : motif_idx + m], axis=1)
+    subseqs = core.z_norm(T[:, subseq_idx : subseq_idx + m], axis=1)
     neighbors = core.z_norm(T[:, nn_idx : nn_idx + m], axis=1)
-    D = np.linalg.norm(motifs - neighbors, axis=1)
+    D = np.linalg.norm(subseqs - neighbors, axis=1)
 
     if discords:
         sorted_idx = D[::-1].argsort(axis=0, kind="mergesort")
@@ -260,9 +260,9 @@ def _get_subspace(T, m, motif_idx, nn_idx, k, include=None, discords=False):
 
     n_bit = 8
     bins = _inverse_norm()
-    disc_motifs = _discretize(motifs[S], bins)
+    disc_subseqs = _discretize(subseqs[S], bins)
     disc_neighbors = _discretize(neighbors[S], bins)
-    n_val = np.unique(disc_motifs - disc_neighbors).shape[0]
+    n_val = np.unique(disc_subseqs - disc_neighbors).shape[0]
     bit_size = n_bit * (T.shape[0] * m * 2 - k * m)
     bit_size = bit_size + k * m * np.log2(n_val) + n_val * n_bit
 
@@ -811,12 +811,12 @@ def mstump(T, m, include=None, discords=False):
     Returns
     -------
     P : ndarray
-        The multi-dimensional matrix profile. Each column of the array corresponds
-        to each matrix profile for a given dimension (i.e., the first column is
-        the 1-D matrix profile and the second column is the 2-D matrix profile).
+        The multi-dimensional matrix profile. Each row of the array corresponds
+        to each matrix profile for a given dimension (i.e., the first row is
+        the 1-D matrix profile and the second row is the 2-D matrix profile).
 
     I : ndarray
-        The multi-dimensional matrix profile index where each column of the array
+        The multi-dimensional matrix profile index where each row of the array
         corresponds to each matrix profile index for a given dimension.
 
     Notes
@@ -874,4 +874,4 @@ def mstump(T, m, include=None, discords=False):
         discords,
     )
 
-    return P.T, I.T
+    return P, I

stumpy/mstump.py

@@ -55,12 +55,12 @@ def mstumped(dask_client, T, m, include=None, discords=False):
     Returns
     -------
     P : ndarray
-        The multi-dimensional matrix profile. Each column of the array corresponds
-        to each matrix profile for a given dimension (i.e., the first column is
-        the 1-D matrix profile and the second column is the 2-D matrix profile).
+        The multi-dimensional matrix profile. Each row of the array corresponds
+        to each matrix profile for a given dimension (i.e., the first row is
+        the 1-D matrix profile and the second row is the 2-D matrix profile).
 
     I : ndarray
-        The multi-dimensional matrix profile index where each column of the array
+        The multi-dimensional matrix profile index where each row of the array
         corresponds to each matrix profile index for a given dimension.
 
     Notes
@@ -167,4 +167,4 @@ def mstumped(dask_client, T, m, include=None, discords=False):
     for future in futures:
         dask_client.cancel(future)
 
-    return P.T, I.T
+    return P, I

stumpy/mstumped.py

@@ -292,7 +292,7 @@ def mstump(T, m, excl_zone, include=None, discords=False):
             P[dim, col_mask] = P_i[dim, col_mask]
             I[dim, col_mask] = I_i[dim, col_mask]
 
-    return P.T, I.T
+    return P, I
 
 
 def subspace(T, m, motif_idx, nn_idx, k, include=None, discords=False):

tests/naive.py

@@ -85,8 +85,8 @@ def test_query_mstump_profile(T, m):
     excl_zone = int(np.ceil(m / 4))
     for query_idx in range(T.shape[0] - m + 1):
         ref_P, ref_I = naive.mstump(T, m, excl_zone)
-        ref_P = ref_P[query_idx, :]
-        ref_I = ref_I[query_idx, :]
+        ref_P = ref_P[:, query_idx]
+        ref_I = ref_I[:, query_idx]
 
         M_T, Σ_T = core.compute_mean_std(T, m)
         comp_P, comp_I = _query_mstump_profile(
@@ -103,8 +103,8 @@ def test_get_first_mstump_profile(T, m):
     start = 0
 
     ref_P, ref_I = naive.mstump(T, m, excl_zone)
-    ref_P = ref_P[start, :]
-    ref_I = ref_I[start, :]
+    ref_P = ref_P[:, start]
+    ref_I = ref_I[:, start]
 
     M_T, Σ_T = core.compute_mean_std(T, m)
     comp_P, comp_I = _get_first_mstump_profile(
@@ -161,29 +161,29 @@ def test_subspace_include(T, m):
 
 @pytest.mark.parametrize("T, m", test_data)
 def test_subspace_discords(T, m):
-    motif_idx = 1
+    discord_idx = 1
     nn_idx = 4
 
     for k in range(T.shape[0]):
-        ref_S = naive.subspace(T, m, motif_idx, nn_idx, k, discords=True)
-        comp_S = _get_subspace(T, m, motif_idx, nn_idx, k, discords=True)
+        ref_S = naive.subspace(T, m, discord_idx, nn_idx, k, discords=True)
+        comp_S = _get_subspace(T, m, discord_idx, nn_idx, k, discords=True)
         npt.assert_almost_equal(ref_S, comp_S)
 
 
 @pytest.mark.parametrize("T, m", test_data)
 def test_subspace_include_discords(T, m):
-    motif_idx = 1
+    discord_idx = 1
     nn_idx = 4
     for width in range(T.shape[0]):
         for i in range(T.shape[0] - width):
             include = np.asarray(range(i, i + width + 1))
 
             for k in range(T.shape[0]):
                 ref_S = naive.subspace(
-                    T, m, motif_idx, nn_idx, k, include, discords=True
+                    T, m, discord_idx, nn_idx, k, include, discords=True
                 )
                 comp_S = _get_subspace(
-                    T, m, motif_idx, nn_idx, k, include, discords=True
+                    T, m, discord_idx, nn_idx, k, include, discords=True
                 )
                 npt.assert_almost_equal(ref_S, comp_S)
 
@@ -195,8 +195,8 @@ def test_naive_mstump():
     zone = int(np.ceil(m / 4))
 
     ref_mp = naive.stamp(T[0], m, exclusion_zone=zone)
-    ref_P = ref_mp[np.newaxis, :, 0].T
-    ref_I = ref_mp[np.newaxis, :, 1].T
+    ref_P = ref_mp[np.newaxis, :, 0]
+    ref_I = ref_mp[np.newaxis, :, 1]
 
     comp_P, comp_I = naive.mstump(T, m, zone)