distance.py
84.4 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
"""
Distance computations (:mod:`scipy.spatial.distance`)
=====================================================
.. sectionauthor:: Damian Eads
Function Reference
------------------
Distance matrix computation from a collection of raw observation vectors
stored in a rectangular array.
.. autosummary::
:toctree: generated/
pdist -- pairwise distances between observation vectors.
cdist -- distances between two collections of observation vectors
squareform -- convert distance matrix to a condensed one and vice versa
directed_hausdorff -- directed Hausdorff distance between arrays
Predicates for checking the validity of distance matrices, both
condensed and redundant. Also contained in this module are functions
for computing the number of observations in a distance matrix.
.. autosummary::
:toctree: generated/
is_valid_dm -- checks for a valid distance matrix
is_valid_y -- checks for a valid condensed distance matrix
num_obs_dm -- # of observations in a distance matrix
num_obs_y -- # of observations in a condensed distance matrix
Distance functions between two numeric vectors ``u`` and ``v``. Computing
distances over a large collection of vectors is inefficient for these
functions. Use ``pdist`` for this purpose.
.. autosummary::
:toctree: generated/
braycurtis -- the Bray-Curtis distance.
canberra -- the Canberra distance.
chebyshev -- the Chebyshev distance.
cityblock -- the Manhattan distance.
correlation -- the Correlation distance.
cosine -- the Cosine distance.
euclidean -- the Euclidean distance.
jensenshannon -- the Jensen-Shannon distance.
mahalanobis -- the Mahalanobis distance.
minkowski -- the Minkowski distance.
seuclidean -- the normalized Euclidean distance.
sqeuclidean -- the squared Euclidean distance.
wminkowski -- (deprecated) alias of `minkowski`.
Distance functions between two boolean vectors (representing sets) ``u`` and
``v``. As in the case of numerical vectors, ``pdist`` is more efficient for
computing the distances between all pairs.
.. autosummary::
:toctree: generated/
dice -- the Dice dissimilarity.
hamming -- the Hamming distance.
jaccard -- the Jaccard distance.
kulsinski -- the Kulsinski distance.
rogerstanimoto -- the Rogers-Tanimoto dissimilarity.
russellrao -- the Russell-Rao dissimilarity.
sokalmichener -- the Sokal-Michener dissimilarity.
sokalsneath -- the Sokal-Sneath dissimilarity.
yule -- the Yule dissimilarity.
:func:`hamming` also operates over discrete numerical vectors.
"""
# Copyright (C) Damian Eads, 2007-2008. New BSD License.
from __future__ import division, print_function, absolute_import
__all__ = [
'braycurtis',
'canberra',
'cdist',
'chebyshev',
'cityblock',
'correlation',
'cosine',
'dice',
'directed_hausdorff',
'euclidean',
'hamming',
'is_valid_dm',
'is_valid_y',
'jaccard',
'jensenshannon',
'kulsinski',
'mahalanobis',
'matching',
'minkowski',
'num_obs_dm',
'num_obs_y',
'pdist',
'rogerstanimoto',
'russellrao',
'seuclidean',
'sokalmichener',
'sokalsneath',
'sqeuclidean',
'squareform',
'wminkowski',
'yule'
]
import warnings
import numpy as np
from functools import partial
from collections import namedtuple
from scipy._lib.six import callable, string_types
from scipy._lib.six import xrange
from scipy._lib._util import _asarray_validated
from . import _distance_wrap
from . import _hausdorff
from ..linalg import norm
from ..special import rel_entr
def _args_to_kwargs_xdist(args, kwargs, metric, func_name):
"""
Convert legacy positional arguments to keyword arguments for pdist/cdist.
"""
if not args:
return kwargs
if (callable(metric) and metric not in [
braycurtis, canberra, chebyshev, cityblock, correlation, cosine,
dice, euclidean, hamming, jaccard, jensenshannon, kulsinski,
mahalanobis, matching, minkowski, rogerstanimoto, russellrao,
seuclidean, sokalmichener, sokalsneath, sqeuclidean, yule,
wminkowski]):
raise TypeError('When using a custom metric arguments must be passed'
'as keyword (i.e., ARGNAME=ARGVALUE)')
if func_name == 'pdist':
old_arg_names = ['p', 'w', 'V', 'VI']
else:
old_arg_names = ['p', 'V', 'VI', 'w']
num_args = len(args)
warnings.warn('%d metric parameters have been passed as positional.'
'This will raise an error in a future version.'
'Please pass arguments as keywords(i.e., ARGNAME=ARGVALUE)'
% num_args, DeprecationWarning)
if num_args > 4:
raise ValueError('Deprecated %s signature accepts only 4'
'positional arguments (%s), %d given.'
% (func_name, ', '.join(old_arg_names), num_args))
for old_arg, arg in zip(old_arg_names, args):
if old_arg in kwargs:
raise TypeError('%s() got multiple values for argument %s'
% (func_name, old_arg))
kwargs[old_arg] = arg
return kwargs
def _copy_array_if_base_present(a):
"""Copy the array if its base points to a parent array."""
if a.base is not None:
return a.copy()
return a
def _correlation_cdist_wrap(XA, XB, dm, **kwargs):
XA = XA - XA.mean(axis=1, keepdims=True)
XB = XB - XB.mean(axis=1, keepdims=True)
_distance_wrap.cdist_cosine_double_wrap(XA, XB, dm, **kwargs)
def _correlation_pdist_wrap(X, dm, **kwargs):
X2 = X - X.mean(axis=1, keepdims=True)
_distance_wrap.pdist_cosine_double_wrap(X2, dm, **kwargs)
def _convert_to_type(X, out_type):
return np.ascontiguousarray(X, dtype=out_type)
def _filter_deprecated_kwargs(kwargs, args_blacklist):
# Filtering out old default keywords
for k in args_blacklist:
if k in kwargs:
del kwargs[k]
warnings.warn('Got unexpected kwarg %s. This will raise an error'
' in a future version.' % k, DeprecationWarning)
def _nbool_correspond_all(u, v, w=None):
if u.dtype == v.dtype == bool and w is None:
not_u = ~u
not_v = ~v
nff = (not_u & not_v).sum()
nft = (not_u & v).sum()
ntf = (u & not_v).sum()
ntt = (u & v).sum()
else:
dtype = np.find_common_type([int], [u.dtype, v.dtype])
u = u.astype(dtype)
v = v.astype(dtype)
not_u = 1.0 - u
not_v = 1.0 - v
if w is not None:
not_u = w * not_u
u = w * u
nff = (not_u * not_v).sum()
nft = (not_u * v).sum()
ntf = (u * not_v).sum()
ntt = (u * v).sum()
return (nff, nft, ntf, ntt)
def _nbool_correspond_ft_tf(u, v, w=None):
if u.dtype == v.dtype == bool and w is None:
not_u = ~u
not_v = ~v
nft = (not_u & v).sum()
ntf = (u & not_v).sum()
else:
dtype = np.find_common_type([int], [u.dtype, v.dtype])
u = u.astype(dtype)
v = v.astype(dtype)
not_u = 1.0 - u
not_v = 1.0 - v
if w is not None:
not_u = w * not_u
u = w * u
nft = (not_u * v).sum()
ntf = (u * not_v).sum()
return (nft, ntf)
def _validate_cdist_input(XA, XB, mA, mB, n, metric_name, **kwargs):
if metric_name is not None:
# get supported types
types = _METRICS[metric_name].types
# choose best type
typ = types[types.index(XA.dtype)] if XA.dtype in types else types[0]
# validate data
XA = _convert_to_type(XA, out_type=typ)
XB = _convert_to_type(XB, out_type=typ)
# validate kwargs
_validate_kwargs = _METRICS[metric_name].validator
if _validate_kwargs:
kwargs = _validate_kwargs(np.vstack([XA, XB]), mA + mB, n, **kwargs)
else:
typ = None
return XA, XB, typ, kwargs
def _validate_hamming_kwargs(X, m, n, **kwargs):
w = kwargs.get('w', np.ones((n,), dtype='double'))
if w.ndim != 1 or w.shape[0] != n:
raise ValueError("Weights must have same size as input vector. %d vs. %d" % (w.shape[0], n))
kwargs['w'] = _validate_weights(w)
return kwargs
def _validate_mahalanobis_kwargs(X, m, n, **kwargs):
VI = kwargs.pop('VI', None)
if VI is None:
if m <= n:
# There are fewer observations than the dimension of
# the observations.
raise ValueError("The number of observations (%d) is too "
"small; the covariance matrix is "
"singular. For observations with %d "
"dimensions, at least %d observations "
"are required." % (m, n, n + 1))
CV = np.atleast_2d(np.cov(X.astype(np.double).T))
VI = np.linalg.inv(CV).T.copy()
kwargs["VI"] = _convert_to_double(VI)
return kwargs
def _validate_minkowski_kwargs(X, m, n, **kwargs):
if 'p' not in kwargs:
kwargs['p'] = 2.
return kwargs
def _validate_pdist_input(X, m, n, metric_name, **kwargs):
if metric_name is not None:
# get supported types
types = _METRICS[metric_name].types
# choose best type
typ = types[types.index(X.dtype)] if X.dtype in types else types[0]
# validate data
X = _convert_to_type(X, out_type=typ)
# validate kwargs
_validate_kwargs = _METRICS[metric_name].validator
if _validate_kwargs:
kwargs = _validate_kwargs(X, m, n, **kwargs)
else:
typ = None
return X, typ, kwargs
def _validate_seuclidean_kwargs(X, m, n, **kwargs):
V = kwargs.pop('V', None)
if V is None:
V = np.var(X.astype(np.double), axis=0, ddof=1)
else:
V = np.asarray(V, order='c')
if V.dtype != np.double:
raise TypeError('Variance vector V must contain doubles.')
if len(V.shape) != 1:
raise ValueError('Variance vector V must '
'be one-dimensional.')
if V.shape[0] != n:
raise ValueError('Variance vector V must be of the same '
'dimension as the vectors on which the distances '
'are computed.')
kwargs['V'] = _convert_to_double(V)
return kwargs
def _validate_vector(u, dtype=None):
# XXX Is order='c' really necessary?
u = np.asarray(u, dtype=dtype, order='c').squeeze()
# Ensure values such as u=1 and u=[1] still return 1-D arrays.
u = np.atleast_1d(u)
if u.ndim > 1:
raise ValueError("Input vector should be 1-D.")
return u
def _validate_weights(w, dtype=np.double):
w = _validate_vector(w, dtype=dtype)
if np.any(w < 0):
raise ValueError("Input weights should be all non-negative")
return w
def _validate_wminkowski_kwargs(X, m, n, **kwargs):
w = kwargs.pop('w', None)
if w is None:
raise ValueError('weighted minkowski requires a weight '
'vector `w` to be given.')
kwargs['w'] = _validate_weights(w)
if 'p' not in kwargs:
kwargs['p'] = 2.
return kwargs
def directed_hausdorff(u, v, seed=0):
"""
Compute the directed Hausdorff distance between two N-D arrays.
Distances between pairs are calculated using a Euclidean metric.
Parameters
----------
u : (M,N) ndarray
Input array.
v : (O,N) ndarray
Input array.
seed : int or None
Local `numpy.random.mtrand.RandomState` seed. Default is 0, a random
shuffling of u and v that guarantees reproducibility.
Returns
-------
d : double
The directed Hausdorff distance between arrays `u` and `v`,
index_1 : int
index of point contributing to Hausdorff pair in `u`
index_2 : int
index of point contributing to Hausdorff pair in `v`
Raises
------
ValueError
An exception is thrown if `u` and `v` do not have
the same number of columns.
Notes
-----
Uses the early break technique and the random sampling approach
described by [1]_. Although worst-case performance is ``O(m * o)``
(as with the brute force algorithm), this is unlikely in practice
as the input data would have to require the algorithm to explore
every single point interaction, and after the algorithm shuffles
the input points at that. The best case performance is O(m), which
is satisfied by selecting an inner loop distance that is less than
cmax and leads to an early break as often as possible. The authors
have formally shown that the average runtime is closer to O(m).
.. versionadded:: 0.19.0
References
----------
.. [1] A. A. Taha and A. Hanbury, "An efficient algorithm for
calculating the exact Hausdorff distance." IEEE Transactions On
Pattern Analysis And Machine Intelligence, vol. 37 pp. 2153-63,
2015.
See Also
--------
scipy.spatial.procrustes : Another similarity test for two data sets
Examples
--------
Find the directed Hausdorff distance between two 2-D arrays of
coordinates:
>>> from scipy.spatial.distance import directed_hausdorff
>>> u = np.array([(1.0, 0.0),
... (0.0, 1.0),
... (-1.0, 0.0),
... (0.0, -1.0)])
>>> v = np.array([(2.0, 0.0),
... (0.0, 2.0),
... (-2.0, 0.0),
... (0.0, -4.0)])
>>> directed_hausdorff(u, v)[0]
2.23606797749979
>>> directed_hausdorff(v, u)[0]
3.0
Find the general (symmetric) Hausdorff distance between two 2-D
arrays of coordinates:
>>> max(directed_hausdorff(u, v)[0], directed_hausdorff(v, u)[0])
3.0
Find the indices of the points that generate the Hausdorff distance
(the Hausdorff pair):
>>> directed_hausdorff(v, u)[1:]
(3, 3)
"""
u = np.asarray(u, dtype=np.float64, order='c')
v = np.asarray(v, dtype=np.float64, order='c')
if u.shape[1] != v.shape[1]:
raise ValueError('u and v need to have the same '
'number of columns')
result = _hausdorff.directed_hausdorff(u, v, seed)
return result
def minkowski(u, v, p=2, w=None):
"""
Compute the Minkowski distance between two 1-D arrays.
The Minkowski distance between 1-D arrays `u` and `v`,
is defined as
.. math::
{||u-v||}_p = (\\sum{|u_i - v_i|^p})^{1/p}.
\\left(\\sum{w_i(|(u_i - v_i)|^p)}\\right)^{1/p}.
Parameters
----------
u : (N,) array_like
Input array.
v : (N,) array_like
Input array.
p : int
The order of the norm of the difference :math:`{||u-v||}_p`.
w : (N,) array_like, optional
The weights for each value in `u` and `v`. Default is None,
which gives each value a weight of 1.0
Returns
-------
minkowski : double
The Minkowski distance between vectors `u` and `v`.
Examples
--------
>>> from scipy.spatial import distance
>>> distance.minkowski([1, 0, 0], [0, 1, 0], 1)
2.0
>>> distance.minkowski([1, 0, 0], [0, 1, 0], 2)
1.4142135623730951
>>> distance.minkowski([1, 0, 0], [0, 1, 0], 3)
1.2599210498948732
>>> distance.minkowski([1, 1, 0], [0, 1, 0], 1)
1.0
>>> distance.minkowski([1, 1, 0], [0, 1, 0], 2)
1.0
>>> distance.minkowski([1, 1, 0], [0, 1, 0], 3)
1.0
"""
u = _validate_vector(u)
v = _validate_vector(v)
if p < 1:
raise ValueError("p must be at least 1")
u_v = u - v
if w is not None:
w = _validate_weights(w)
if p == 1:
root_w = w
if p == 2:
# better precision and speed
root_w = np.sqrt(w)
else:
root_w = np.power(w, 1/p)
u_v = root_w * u_v
dist = norm(u_v, ord=p)
return dist
# `minkowski` gained weights in scipy 1.0. Once we're at say version 1.3,
# deprecated `wminkowski`. Not done at once because it would be annoying for
# downstream libraries that used `wminkowski` and support multiple scipy
# versions.
def wminkowski(u, v, p, w):
"""
Compute the weighted Minkowski distance between two 1-D arrays.
The weighted Minkowski distance between `u` and `v`, defined as
.. math::
\\left(\\sum{(|w_i (u_i - v_i)|^p)}\\right)^{1/p}.
Parameters
----------
u : (N,) array_like
Input array.
v : (N,) array_like
Input array.
p : int
The order of the norm of the difference :math:`{||u-v||}_p`.
w : (N,) array_like
The weight vector.
Returns
-------
wminkowski : double
The weighted Minkowski distance between vectors `u` and `v`.
Notes
-----
`wminkowski` is DEPRECATED. It implements a definition where weights
are powered. It is recommended to use the weighted version of `minkowski`
instead. This function will be removed in a future version of scipy.
Examples
--------
>>> from scipy.spatial import distance
>>> distance.wminkowski([1, 0, 0], [0, 1, 0], 1, np.ones(3))
2.0
>>> distance.wminkowski([1, 0, 0], [0, 1, 0], 2, np.ones(3))
1.4142135623730951
>>> distance.wminkowski([1, 0, 0], [0, 1, 0], 3, np.ones(3))
1.2599210498948732
>>> distance.wminkowski([1, 1, 0], [0, 1, 0], 1, np.ones(3))
1.0
>>> distance.wminkowski([1, 1, 0], [0, 1, 0], 2, np.ones(3))
1.0
>>> distance.wminkowski([1, 1, 0], [0, 1, 0], 3, np.ones(3))
1.0
"""
w = _validate_weights(w)
return minkowski(u, v, p=p, w=w**p)
def euclidean(u, v, w=None):
"""
Computes the Euclidean distance between two 1-D arrays.
The Euclidean distance between 1-D arrays `u` and `v`, is defined as
.. math::
{||u-v||}_2
\\left(\\sum{(w_i |(u_i - v_i)|^2)}\\right)^{1/2}
Parameters
----------
u : (N,) array_like
Input array.
v : (N,) array_like
Input array.
w : (N,) array_like, optional
The weights for each value in `u` and `v`. Default is None,
which gives each value a weight of 1.0
Returns
-------
euclidean : double
The Euclidean distance between vectors `u` and `v`.
Examples
--------
>>> from scipy.spatial import distance
>>> distance.euclidean([1, 0, 0], [0, 1, 0])
1.4142135623730951
>>> distance.euclidean([1, 1, 0], [0, 1, 0])
1.0
"""
return minkowski(u, v, p=2, w=w)
def sqeuclidean(u, v, w=None):
"""
Compute the squared Euclidean distance between two 1-D arrays.
The squared Euclidean distance between `u` and `v` is defined as
.. math::
{||u-v||}_2^2
\\left(\\sum{(w_i |(u_i - v_i)|^2)}\\right)
Parameters
----------
u : (N,) array_like
Input array.
v : (N,) array_like
Input array.
w : (N,) array_like, optional
The weights for each value in `u` and `v`. Default is None,
which gives each value a weight of 1.0
Returns
-------
sqeuclidean : double
The squared Euclidean distance between vectors `u` and `v`.
Examples
--------
>>> from scipy.spatial import distance
>>> distance.sqeuclidean([1, 0, 0], [0, 1, 0])
2.0
>>> distance.sqeuclidean([1, 1, 0], [0, 1, 0])
1.0
"""
# Preserve float dtypes, but convert everything else to np.float64
# for stability.
utype, vtype = None, None
if not (hasattr(u, "dtype") and np.issubdtype(u.dtype, np.inexact)):
utype = np.float64
if not (hasattr(v, "dtype") and np.issubdtype(v.dtype, np.inexact)):
vtype = np.float64
u = _validate_vector(u, dtype=utype)
v = _validate_vector(v, dtype=vtype)
u_v = u - v
u_v_w = u_v # only want weights applied once
if w is not None:
w = _validate_weights(w)
u_v_w = w * u_v
return np.dot(u_v, u_v_w)
def correlation(u, v, w=None, centered=True):
"""
Compute the correlation distance between two 1-D arrays.
The correlation distance between `u` and `v`, is
defined as
.. math::
1 - \\frac{(u - \\bar{u}) \\cdot (v - \\bar{v})}
{{||(u - \\bar{u})||}_2 {||(v - \\bar{v})||}_2}
where :math:`\\bar{u}` is the mean of the elements of `u`
and :math:`x \\cdot y` is the dot product of :math:`x` and :math:`y`.
Parameters
----------
u : (N,) array_like
Input array.
v : (N,) array_like
Input array.
w : (N,) array_like, optional
The weights for each value in `u` and `v`. Default is None,
which gives each value a weight of 1.0
Returns
-------
correlation : double
The correlation distance between 1-D array `u` and `v`.
"""
u = _validate_vector(u)
v = _validate_vector(v)
if w is not None:
w = _validate_weights(w)
if centered:
umu = np.average(u, weights=w)
vmu = np.average(v, weights=w)
u = u - umu
v = v - vmu
uv = np.average(u * v, weights=w)
uu = np.average(np.square(u), weights=w)
vv = np.average(np.square(v), weights=w)
dist = 1.0 - uv / np.sqrt(uu * vv)
return dist
def cosine(u, v, w=None):
"""
Compute the Cosine distance between 1-D arrays.
The Cosine distance between `u` and `v`, is defined as
.. math::
1 - \\frac{u \\cdot v}
{||u||_2 ||v||_2}.
where :math:`u \\cdot v` is the dot product of :math:`u` and
:math:`v`.
Parameters
----------
u : (N,) array_like
Input array.
v : (N,) array_like
Input array.
w : (N,) array_like, optional
The weights for each value in `u` and `v`. Default is None,
which gives each value a weight of 1.0
Returns
-------
cosine : double
The Cosine distance between vectors `u` and `v`.
Examples
--------
>>> from scipy.spatial import distance
>>> distance.cosine([1, 0, 0], [0, 1, 0])
1.0
>>> distance.cosine([100, 0, 0], [0, 1, 0])
1.0
>>> distance.cosine([1, 1, 0], [0, 1, 0])
0.29289321881345254
"""
# cosine distance is also referred to as 'uncentered correlation',
# or 'reflective correlation'
return correlation(u, v, w=w, centered=False)
def hamming(u, v, w=None):
"""
Compute the Hamming distance between two 1-D arrays.
The Hamming distance between 1-D arrays `u` and `v`, is simply the
proportion of disagreeing components in `u` and `v`. If `u` and `v` are
boolean vectors, the Hamming distance is
.. math::
\\frac{c_{01} + c_{10}}{n}
where :math:`c_{ij}` is the number of occurrences of
:math:`\\mathtt{u[k]} = i` and :math:`\\mathtt{v[k]} = j` for
:math:`k < n`.
Parameters
----------
u : (N,) array_like
Input array.
v : (N,) array_like
Input array.
w : (N,) array_like, optional
The weights for each value in `u` and `v`. Default is None,
which gives each value a weight of 1.0
Returns
-------
hamming : double
The Hamming distance between vectors `u` and `v`.
Examples
--------
>>> from scipy.spatial import distance
>>> distance.hamming([1, 0, 0], [0, 1, 0])
0.66666666666666663
>>> distance.hamming([1, 0, 0], [1, 1, 0])
0.33333333333333331
>>> distance.hamming([1, 0, 0], [2, 0, 0])
0.33333333333333331
>>> distance.hamming([1, 0, 0], [3, 0, 0])
0.33333333333333331
"""
u = _validate_vector(u)
v = _validate_vector(v)
if u.shape != v.shape:
raise ValueError('The 1d arrays must have equal lengths.')
u_ne_v = u != v
if w is not None:
w = _validate_weights(w)
return np.average(u_ne_v, weights=w)
def jaccard(u, v, w=None):
"""
Compute the Jaccard-Needham dissimilarity between two boolean 1-D arrays.
The Jaccard-Needham dissimilarity between 1-D boolean arrays `u` and `v`,
is defined as
.. math::
\\frac{c_{TF} + c_{FT}}
{c_{TT} + c_{FT} + c_{TF}}
where :math:`c_{ij}` is the number of occurrences of
:math:`\\mathtt{u[k]} = i` and :math:`\\mathtt{v[k]} = j` for
:math:`k < n`.
Parameters
----------
u : (N,) array_like, bool
Input array.
v : (N,) array_like, bool
Input array.
w : (N,) array_like, optional
The weights for each value in `u` and `v`. Default is None,
which gives each value a weight of 1.0
Returns
-------
jaccard : double
The Jaccard distance between vectors `u` and `v`.
Notes
-----
When both `u` and `v` lead to a `0/0` division i.e. there is no overlap
between the items in the vectors the returned distance is 0. See the
Wikipedia page on the Jaccard index [1]_, and this paper [2]_.
.. versionchanged:: 1.2.0
Previously, when `u` and `v` lead to a `0/0` division, the function
would return NaN. This was changed to return 0 instead.
References
----------
.. [1] https://en.wikipedia.org/wiki/Jaccard_index
.. [2] S. Kosub, "A note on the triangle inequality for the Jaccard
distance", 2016, Available online: https://arxiv.org/pdf/1612.02696.pdf
Examples
--------
>>> from scipy.spatial import distance
>>> distance.jaccard([1, 0, 0], [0, 1, 0])
1.0
>>> distance.jaccard([1, 0, 0], [1, 1, 0])
0.5
>>> distance.jaccard([1, 0, 0], [1, 2, 0])
0.5
>>> distance.jaccard([1, 0, 0], [1, 1, 1])
0.66666666666666663
"""
u = _validate_vector(u)
v = _validate_vector(v)
nonzero = np.bitwise_or(u != 0, v != 0)
unequal_nonzero = np.bitwise_and((u != v), nonzero)
if w is not None:
w = _validate_weights(w)
nonzero = w * nonzero
unequal_nonzero = w * unequal_nonzero
a = np.double(unequal_nonzero.sum())
b = np.double(nonzero.sum())
return (a / b) if b != 0 else 0
def kulsinski(u, v, w=None):
"""
Compute the Kulsinski dissimilarity between two boolean 1-D arrays.
The Kulsinski dissimilarity between two boolean 1-D arrays `u` and `v`,
is defined as
.. math::
\\frac{c_{TF} + c_{FT} - c_{TT} + n}
{c_{FT} + c_{TF} + n}
where :math:`c_{ij}` is the number of occurrences of
:math:`\\mathtt{u[k]} = i` and :math:`\\mathtt{v[k]} = j` for
:math:`k < n`.
Parameters
----------
u : (N,) array_like, bool
Input array.
v : (N,) array_like, bool
Input array.
w : (N,) array_like, optional
The weights for each value in `u` and `v`. Default is None,
which gives each value a weight of 1.0
Returns
-------
kulsinski : double
The Kulsinski distance between vectors `u` and `v`.
Examples
--------
>>> from scipy.spatial import distance
>>> distance.kulsinski([1, 0, 0], [0, 1, 0])
1.0
>>> distance.kulsinski([1, 0, 0], [1, 1, 0])
0.75
>>> distance.kulsinski([1, 0, 0], [2, 1, 0])
0.33333333333333331
>>> distance.kulsinski([1, 0, 0], [3, 1, 0])
-0.5
"""
u = _validate_vector(u)
v = _validate_vector(v)
if w is None:
n = float(len(u))
else:
w = _validate_weights(w)
n = w.sum()
(nff, nft, ntf, ntt) = _nbool_correspond_all(u, v, w=w)
return (ntf + nft - ntt + n) / (ntf + nft + n)
def seuclidean(u, v, V):
"""
Return the standardized Euclidean distance between two 1-D arrays.
The standardized Euclidean distance between `u` and `v`.
Parameters
----------
u : (N,) array_like
Input array.
v : (N,) array_like
Input array.
V : (N,) array_like
`V` is an 1-D array of component variances. It is usually computed
among a larger collection vectors.
Returns
-------
seuclidean : double
The standardized Euclidean distance between vectors `u` and `v`.
Examples
--------
>>> from scipy.spatial import distance
>>> distance.seuclidean([1, 0, 0], [0, 1, 0], [0.1, 0.1, 0.1])
4.4721359549995796
>>> distance.seuclidean([1, 0, 0], [0, 1, 0], [1, 0.1, 0.1])
3.3166247903553998
>>> distance.seuclidean([1, 0, 0], [0, 1, 0], [10, 0.1, 0.1])
3.1780497164141406
"""
u = _validate_vector(u)
v = _validate_vector(v)
V = _validate_vector(V, dtype=np.float64)
if V.shape[0] != u.shape[0] or u.shape[0] != v.shape[0]:
raise TypeError('V must be a 1-D array of the same dimension '
'as u and v.')
return euclidean(u, v, w=1/V)
def cityblock(u, v, w=None):
"""
Compute the City Block (Manhattan) distance.
Computes the Manhattan distance between two 1-D arrays `u` and `v`,
which is defined as
.. math::
\\sum_i {\\left| u_i - v_i \\right|}.
Parameters
----------
u : (N,) array_like
Input array.
v : (N,) array_like
Input array.
w : (N,) array_like, optional
The weights for each value in `u` and `v`. Default is None,
which gives each value a weight of 1.0
Returns
-------
cityblock : double
The City Block (Manhattan) distance between vectors `u` and `v`.
Examples
--------
>>> from scipy.spatial import distance
>>> distance.cityblock([1, 0, 0], [0, 1, 0])
2
>>> distance.cityblock([1, 0, 0], [0, 2, 0])
3
>>> distance.cityblock([1, 0, 0], [1, 1, 0])
1
"""
u = _validate_vector(u)
v = _validate_vector(v)
l1_diff = abs(u - v)
if w is not None:
w = _validate_weights(w)
l1_diff = w * l1_diff
return l1_diff.sum()
def mahalanobis(u, v, VI):
"""
Compute the Mahalanobis distance between two 1-D arrays.
The Mahalanobis distance between 1-D arrays `u` and `v`, is defined as
.. math::
\\sqrt{ (u-v) V^{-1} (u-v)^T }
where ``V`` is the covariance matrix. Note that the argument `VI`
is the inverse of ``V``.
Parameters
----------
u : (N,) array_like
Input array.
v : (N,) array_like
Input array.
VI : ndarray
The inverse of the covariance matrix.
Returns
-------
mahalanobis : double
The Mahalanobis distance between vectors `u` and `v`.
Examples
--------
>>> from scipy.spatial import distance
>>> iv = [[1, 0.5, 0.5], [0.5, 1, 0.5], [0.5, 0.5, 1]]
>>> distance.mahalanobis([1, 0, 0], [0, 1, 0], iv)
1.0
>>> distance.mahalanobis([0, 2, 0], [0, 1, 0], iv)
1.0
>>> distance.mahalanobis([2, 0, 0], [0, 1, 0], iv)
1.7320508075688772
"""
u = _validate_vector(u)
v = _validate_vector(v)
VI = np.atleast_2d(VI)
delta = u - v
m = np.dot(np.dot(delta, VI), delta)
return np.sqrt(m)
def chebyshev(u, v, w=None):
"""
Compute the Chebyshev distance.
Computes the Chebyshev distance between two 1-D arrays `u` and `v`,
which is defined as
.. math::
\\max_i {|u_i-v_i|}.
Parameters
----------
u : (N,) array_like
Input vector.
v : (N,) array_like
Input vector.
w : (N,) array_like, optional
The weights for each value in `u` and `v`. Default is None,
which gives each value a weight of 1.0
Returns
-------
chebyshev : double
The Chebyshev distance between vectors `u` and `v`.
Examples
--------
>>> from scipy.spatial import distance
>>> distance.chebyshev([1, 0, 0], [0, 1, 0])
1
>>> distance.chebyshev([1, 1, 0], [0, 1, 0])
1
"""
u = _validate_vector(u)
v = _validate_vector(v)
if w is not None:
w = _validate_weights(w)
has_weight = w > 0
if has_weight.sum() < w.size:
u = u[has_weight]
v = v[has_weight]
return max(abs(u - v))
def braycurtis(u, v, w=None):
"""
Compute the Bray-Curtis distance between two 1-D arrays.
Bray-Curtis distance is defined as
.. math::
\\sum{|u_i-v_i|} / \\sum{|u_i+v_i|}
The Bray-Curtis distance is in the range [0, 1] if all coordinates are
positive, and is undefined if the inputs are of length zero.
Parameters
----------
u : (N,) array_like
Input array.
v : (N,) array_like
Input array.
w : (N,) array_like, optional
The weights for each value in `u` and `v`. Default is None,
which gives each value a weight of 1.0
Returns
-------
braycurtis : double
The Bray-Curtis distance between 1-D arrays `u` and `v`.
Examples
--------
>>> from scipy.spatial import distance
>>> distance.braycurtis([1, 0, 0], [0, 1, 0])
1.0
>>> distance.braycurtis([1, 1, 0], [0, 1, 0])
0.33333333333333331
"""
u = _validate_vector(u)
v = _validate_vector(v, dtype=np.float64)
l1_diff = abs(u - v)
l1_sum = abs(u + v)
if w is not None:
w = _validate_weights(w)
l1_diff = w * l1_diff
l1_sum = w * l1_sum
return l1_diff.sum() / l1_sum.sum()
def canberra(u, v, w=None):
"""
Compute the Canberra distance between two 1-D arrays.
The Canberra distance is defined as
.. math::
d(u,v) = \\sum_i \\frac{|u_i-v_i|}
{|u_i|+|v_i|}.
Parameters
----------
u : (N,) array_like
Input array.
v : (N,) array_like
Input array.
w : (N,) array_like, optional
The weights for each value in `u` and `v`. Default is None,
which gives each value a weight of 1.0
Returns
-------
canberra : double
The Canberra distance between vectors `u` and `v`.
Notes
-----
When `u[i]` and `v[i]` are 0 for given i, then the fraction 0/0 = 0 is
used in the calculation.
Examples
--------
>>> from scipy.spatial import distance
>>> distance.canberra([1, 0, 0], [0, 1, 0])
2.0
>>> distance.canberra([1, 1, 0], [0, 1, 0])
1.0
"""
u = _validate_vector(u)
v = _validate_vector(v, dtype=np.float64)
if w is not None:
w = _validate_weights(w)
olderr = np.seterr(invalid='ignore')
try:
abs_uv = abs(u - v)
abs_u = abs(u)
abs_v = abs(v)
d = abs_uv / (abs_u + abs_v)
if w is not None:
d = w * d
d = np.nansum(d)
finally:
np.seterr(**olderr)
return d
def jensenshannon(p, q, base=None):
"""
Compute the Jensen-Shannon distance (metric) between
two 1-D probability arrays. This is the square root
of the Jensen-Shannon divergence.
The Jensen-Shannon distance between two probability
vectors `p` and `q` is defined as,
.. math::
\\sqrt{\\frac{D(p \\parallel m) + D(q \\parallel m)}{2}}
where :math:`m` is the pointwise mean of :math:`p` and :math:`q`
and :math:`D` is the Kullback-Leibler divergence.
This routine will normalize `p` and `q` if they don't sum to 1.0.
Parameters
----------
p : (N,) array_like
left probability vector
q : (N,) array_like
right probability vector
base : double, optional
the base of the logarithm used to compute the output
if not given, then the routine uses the default base of
scipy.stats.entropy.
Returns
-------
js : double
The Jensen-Shannon distance between `p` and `q`
.. versionadded:: 1.2.0
Examples
--------
>>> from scipy.spatial import distance
>>> distance.jensenshannon([1.0, 0.0, 0.0], [0.0, 1.0, 0.0], 2.0)
1.0
>>> distance.jensenshannon([1.0, 0.0], [0.5, 0.5])
0.46450140402245893
>>> distance.jensenshannon([1.0, 0.0, 0.0], [1.0, 0.0, 0.0])
0.0
"""
p = np.asarray(p)
q = np.asarray(q)
p = p / np.sum(p, axis=0)
q = q / np.sum(q, axis=0)
m = (p + q) / 2.0
left = rel_entr(p, m)
right = rel_entr(q, m)
js = np.sum(left, axis=0) + np.sum(right, axis=0)
if base is not None:
js /= np.log(base)
return np.sqrt(js / 2.0)
def yule(u, v, w=None):
"""
Compute the Yule dissimilarity between two boolean 1-D arrays.
The Yule dissimilarity is defined as
.. math::
\\frac{R}{c_{TT} * c_{FF} + \\frac{R}{2}}
where :math:`c_{ij}` is the number of occurrences of
:math:`\\mathtt{u[k]} = i` and :math:`\\mathtt{v[k]} = j` for
:math:`k < n` and :math:`R = 2.0 * c_{TF} * c_{FT}`.
Parameters
----------
u : (N,) array_like, bool
Input array.
v : (N,) array_like, bool
Input array.
w : (N,) array_like, optional
The weights for each value in `u` and `v`. Default is None,
which gives each value a weight of 1.0
Returns
-------
yule : double
The Yule dissimilarity between vectors `u` and `v`.
Examples
--------
>>> from scipy.spatial import distance
>>> distance.yule([1, 0, 0], [0, 1, 0])
2.0
>>> distance.yule([1, 1, 0], [0, 1, 0])
0.0
"""
u = _validate_vector(u)
v = _validate_vector(v)
if w is not None:
w = _validate_weights(w)
(nff, nft, ntf, ntt) = _nbool_correspond_all(u, v, w=w)
return float(2.0 * ntf * nft / np.array(ntt * nff + ntf * nft))
@np.deprecate(message="spatial.distance.matching is deprecated in scipy 1.0.0; "
"use spatial.distance.hamming instead.")
def matching(u, v, w=None):
"""
Compute the Hamming distance between two boolean 1-D arrays.
This is a deprecated synonym for :func:`hamming`.
"""
return hamming(u, v, w=w)
def dice(u, v, w=None):
"""
Compute the Dice dissimilarity between two boolean 1-D arrays.
The Dice dissimilarity between `u` and `v`, is
.. math::
\\frac{c_{TF} + c_{FT}}
{2c_{TT} + c_{FT} + c_{TF}}
where :math:`c_{ij}` is the number of occurrences of
:math:`\\mathtt{u[k]} = i` and :math:`\\mathtt{v[k]} = j` for
:math:`k < n`.
Parameters
----------
u : (N,) ndarray, bool
Input 1-D array.
v : (N,) ndarray, bool
Input 1-D array.
w : (N,) array_like, optional
The weights for each value in `u` and `v`. Default is None,
which gives each value a weight of 1.0
Returns
-------
dice : double
The Dice dissimilarity between 1-D arrays `u` and `v`.
Examples
--------
>>> from scipy.spatial import distance
>>> distance.dice([1, 0, 0], [0, 1, 0])
1.0
>>> distance.dice([1, 0, 0], [1, 1, 0])
0.3333333333333333
>>> distance.dice([1, 0, 0], [2, 0, 0])
-0.3333333333333333
"""
u = _validate_vector(u)
v = _validate_vector(v)
if w is not None:
w = _validate_weights(w)
if u.dtype == v.dtype == bool and w is None:
ntt = (u & v).sum()
else:
dtype = np.find_common_type([int], [u.dtype, v.dtype])
u = u.astype(dtype)
v = v.astype(dtype)
if w is None:
ntt = (u * v).sum()
else:
ntt = (u * v * w).sum()
(nft, ntf) = _nbool_correspond_ft_tf(u, v, w=w)
return float((ntf + nft) / np.array(2.0 * ntt + ntf + nft))
def rogerstanimoto(u, v, w=None):
"""
Compute the Rogers-Tanimoto dissimilarity between two boolean 1-D arrays.
The Rogers-Tanimoto dissimilarity between two boolean 1-D arrays
`u` and `v`, is defined as
.. math::
\\frac{R}
{c_{TT} + c_{FF} + R}
where :math:`c_{ij}` is the number of occurrences of
:math:`\\mathtt{u[k]} = i` and :math:`\\mathtt{v[k]} = j` for
:math:`k < n` and :math:`R = 2(c_{TF} + c_{FT})`.
Parameters
----------
u : (N,) array_like, bool
Input array.
v : (N,) array_like, bool
Input array.
w : (N,) array_like, optional
The weights for each value in `u` and `v`. Default is None,
which gives each value a weight of 1.0
Returns
-------
rogerstanimoto : double
The Rogers-Tanimoto dissimilarity between vectors
`u` and `v`.
Examples
--------
>>> from scipy.spatial import distance
>>> distance.rogerstanimoto([1, 0, 0], [0, 1, 0])
0.8
>>> distance.rogerstanimoto([1, 0, 0], [1, 1, 0])
0.5
>>> distance.rogerstanimoto([1, 0, 0], [2, 0, 0])
-1.0
"""
u = _validate_vector(u)
v = _validate_vector(v)
if w is not None:
w = _validate_weights(w)
(nff, nft, ntf, ntt) = _nbool_correspond_all(u, v, w=w)
return float(2.0 * (ntf + nft)) / float(ntt + nff + (2.0 * (ntf + nft)))
def russellrao(u, v, w=None):
"""
Compute the Russell-Rao dissimilarity between two boolean 1-D arrays.
The Russell-Rao dissimilarity between two boolean 1-D arrays, `u` and
`v`, is defined as
.. math::
\\frac{n - c_{TT}}
{n}
where :math:`c_{ij}` is the number of occurrences of
:math:`\\mathtt{u[k]} = i` and :math:`\\mathtt{v[k]} = j` for
:math:`k < n`.
Parameters
----------
u : (N,) array_like, bool
Input array.
v : (N,) array_like, bool
Input array.
w : (N,) array_like, optional
The weights for each value in `u` and `v`. Default is None,
which gives each value a weight of 1.0
Returns
-------
russellrao : double
The Russell-Rao dissimilarity between vectors `u` and `v`.
Examples
--------
>>> from scipy.spatial import distance
>>> distance.russellrao([1, 0, 0], [0, 1, 0])
1.0
>>> distance.russellrao([1, 0, 0], [1, 1, 0])
0.6666666666666666
>>> distance.russellrao([1, 0, 0], [2, 0, 0])
0.3333333333333333
"""
u = _validate_vector(u)
v = _validate_vector(v)
if u.dtype == v.dtype == bool and w is None:
ntt = (u & v).sum()
n = float(len(u))
elif w is None:
ntt = (u * v).sum()
n = float(len(u))
else:
w = _validate_weights(w)
ntt = (u * v * w).sum()
n = w.sum()
return float(n - ntt) / n
def sokalmichener(u, v, w=None):
"""
Compute the Sokal-Michener dissimilarity between two boolean 1-D arrays.
The Sokal-Michener dissimilarity between boolean 1-D arrays `u` and `v`,
is defined as
.. math::
\\frac{R}
{S + R}
where :math:`c_{ij}` is the number of occurrences of
:math:`\\mathtt{u[k]} = i` and :math:`\\mathtt{v[k]} = j` for
:math:`k < n`, :math:`R = 2 * (c_{TF} + c_{FT})` and
:math:`S = c_{FF} + c_{TT}`.
Parameters
----------
u : (N,) array_like, bool
Input array.
v : (N,) array_like, bool
Input array.
w : (N,) array_like, optional
The weights for each value in `u` and `v`. Default is None,
which gives each value a weight of 1.0
Returns
-------
sokalmichener : double
The Sokal-Michener dissimilarity between vectors `u` and `v`.
Examples
--------
>>> from scipy.spatial import distance
>>> distance.sokalmichener([1, 0, 0], [0, 1, 0])
0.8
>>> distance.sokalmichener([1, 0, 0], [1, 1, 0])
0.5
>>> distance.sokalmichener([1, 0, 0], [2, 0, 0])
-1.0
"""
u = _validate_vector(u)
v = _validate_vector(v)
if u.dtype == v.dtype == bool and w is None:
ntt = (u & v).sum()
nff = (~u & ~v).sum()
elif w is None:
ntt = (u * v).sum()
nff = ((1.0 - u) * (1.0 - v)).sum()
else:
w = _validate_weights(w)
ntt = (u * v * w).sum()
nff = ((1.0 - u) * (1.0 - v) * w).sum()
(nft, ntf) = _nbool_correspond_ft_tf(u, v)
return float(2.0 * (ntf + nft)) / float(ntt + nff + 2.0 * (ntf + nft))
def sokalsneath(u, v, w=None):
"""
Compute the Sokal-Sneath dissimilarity between two boolean 1-D arrays.
The Sokal-Sneath dissimilarity between `u` and `v`,
.. math::
\\frac{R}
{c_{TT} + R}
where :math:`c_{ij}` is the number of occurrences of
:math:`\\mathtt{u[k]} = i` and :math:`\\mathtt{v[k]} = j` for
:math:`k < n` and :math:`R = 2(c_{TF} + c_{FT})`.
Parameters
----------
u : (N,) array_like, bool
Input array.
v : (N,) array_like, bool
Input array.
w : (N,) array_like, optional
The weights for each value in `u` and `v`. Default is None,
which gives each value a weight of 1.0
Returns
-------
sokalsneath : double
The Sokal-Sneath dissimilarity between vectors `u` and `v`.
Examples
--------
>>> from scipy.spatial import distance
>>> distance.sokalsneath([1, 0, 0], [0, 1, 0])
1.0
>>> distance.sokalsneath([1, 0, 0], [1, 1, 0])
0.66666666666666663
>>> distance.sokalsneath([1, 0, 0], [2, 1, 0])
0.0
>>> distance.sokalsneath([1, 0, 0], [3, 1, 0])
-2.0
"""
u = _validate_vector(u)
v = _validate_vector(v)
if u.dtype == v.dtype == bool and w is None:
ntt = (u & v).sum()
elif w is None:
ntt = (u * v).sum()
else:
w = _validate_weights(w)
ntt = (u * v * w).sum()
(nft, ntf) = _nbool_correspond_ft_tf(u, v, w=w)
denom = np.array(ntt + 2.0 * (ntf + nft))
if not denom.any():
raise ValueError('Sokal-Sneath dissimilarity is not defined for '
'vectors that are entirely false.')
return float(2.0 * (ntf + nft)) / denom
_convert_to_double = partial(_convert_to_type, out_type=np.double)
_convert_to_bool = partial(_convert_to_type, out_type=bool)
# adding python-only wrappers to _distance_wrap module
_distance_wrap.pdist_correlation_double_wrap = _correlation_pdist_wrap
_distance_wrap.cdist_correlation_double_wrap = _correlation_cdist_wrap
# Registry of implemented metrics:
# Dictionary with the following structure:
# {
# metric_name : MetricInfo(aka, types=[double], validator=None)
# }
#
# Where:
# `metric_name` must be equal to python metric name
#
# MetricInfo is a named tuple with fields:
# 'aka' : [list of aliases],
#
# 'validator': f(X, m, n, **kwargs) # function that check kwargs and
# # computes default values.
#
# 'types': [list of supported types], # X (pdist) and XA (cdist) are used to
# # choose the type. if there is no match
# # the first type is used. Default double
# }
MetricInfo = namedtuple("MetricInfo", 'aka types validator ')
MetricInfo.__new__.__defaults__ = (['double'], None)
_METRICS = {
'braycurtis': MetricInfo(aka=['braycurtis']),
'canberra': MetricInfo(aka=['canberra']),
'chebyshev': MetricInfo(aka=['chebychev', 'chebyshev', 'cheby', 'cheb', 'ch']),
'cityblock': MetricInfo(aka=['cityblock', 'cblock', 'cb', 'c']),
'correlation': MetricInfo(aka=['correlation', 'co']),
'cosine': MetricInfo(aka=['cosine', 'cos']),
'dice': MetricInfo(aka=['dice'], types=['bool']),
'euclidean': MetricInfo(aka=['euclidean', 'euclid', 'eu', 'e']),
'hamming': MetricInfo(aka=['matching', 'hamming', 'hamm', 'ha', 'h'],
types=['double', 'bool'],
validator=_validate_hamming_kwargs),
'jaccard': MetricInfo(aka=['jaccard', 'jacc', 'ja', 'j'],
types=['double', 'bool']),
'jensenshannon': MetricInfo(aka=['jensenshannon', 'js'],
types=['double']),
'kulsinski': MetricInfo(aka=['kulsinski'], types=['bool']),
'mahalanobis': MetricInfo(aka=['mahalanobis', 'mahal', 'mah'],
validator=_validate_mahalanobis_kwargs),
'minkowski': MetricInfo(aka=['minkowski', 'mi', 'm', 'pnorm'],
validator=_validate_minkowski_kwargs),
'rogerstanimoto': MetricInfo(aka=['rogerstanimoto'], types=['bool']),
'russellrao': MetricInfo(aka=['russellrao'], types=['bool']),
'seuclidean': MetricInfo(aka=['seuclidean', 'se', 's'],
validator=_validate_seuclidean_kwargs),
'sokalmichener': MetricInfo(aka=['sokalmichener'], types=['bool']),
'sokalsneath': MetricInfo(aka=['sokalsneath'], types=['bool']),
'sqeuclidean': MetricInfo(aka=['sqeuclidean', 'sqe', 'sqeuclid']),
'wminkowski': MetricInfo(aka=['wminkowski', 'wmi', 'wm', 'wpnorm'],
validator=_validate_wminkowski_kwargs),
'yule': MetricInfo(aka=['yule'], types=['bool']),
}
_METRIC_ALIAS = dict((alias, name)
for name, info in _METRICS.items()
for alias in info.aka)
_METRICS_NAMES = list(_METRICS.keys())
_TEST_METRICS = {'test_' + name: globals()[name] for name in _METRICS.keys()}
def _select_weighted_metric(mstr, kwargs, out):
kwargs = dict(kwargs)
if "w" in kwargs and kwargs["w"] is None:
# w=None is the same as omitting it
kwargs.pop("w")
if mstr.startswith("test_") or mstr in _METRICS['wminkowski'].aka + _METRICS['hamming'].aka:
# These support weights
pass
elif "w" in kwargs:
if (mstr in _METRICS['seuclidean'].aka or
mstr in _METRICS['mahalanobis'].aka):
raise ValueError("metric %s incompatible with weights" % mstr)
# XXX: C-versions do not support weights
# need to use python version for weighting
kwargs['out'] = out
mstr = "test_%s" % mstr
return mstr, kwargs
def pdist(X, metric='euclidean', *args, **kwargs):
"""
Pairwise distances between observations in n-dimensional space.
See Notes for common calling conventions.
Parameters
----------
X : ndarray
An m by n array of m original observations in an
n-dimensional space.
metric : str or function, optional
The distance metric to use. The distance function can
be 'braycurtis', 'canberra', 'chebyshev', 'cityblock',
'correlation', 'cosine', 'dice', 'euclidean', 'hamming',
'jaccard', 'jensenshannon', 'kulsinski', 'mahalanobis', 'matching',
'minkowski', 'rogerstanimoto', 'russellrao', 'seuclidean',
'sokalmichener', 'sokalsneath', 'sqeuclidean', 'yule'.
*args : tuple. Deprecated.
Additional arguments should be passed as keyword arguments
**kwargs : dict, optional
Extra arguments to `metric`: refer to each metric documentation for a
list of all possible arguments.
Some possible arguments:
p : scalar
The p-norm to apply for Minkowski, weighted and unweighted.
Default: 2.
w : ndarray
The weight vector for metrics that support weights (e.g., Minkowski).
V : ndarray
The variance vector for standardized Euclidean.
Default: var(X, axis=0, ddof=1)
VI : ndarray
The inverse of the covariance matrix for Mahalanobis.
Default: inv(cov(X.T)).T
out : ndarray.
The output array
If not None, condensed distance matrix Y is stored in this array.
Note: metric independent, it will become a regular keyword arg in a
future scipy version
Returns
-------
Y : ndarray
Returns a condensed distance matrix Y. For
each :math:`i` and :math:`j` (where :math:`i<j<m`),where m is the number
of original observations. The metric ``dist(u=X[i], v=X[j])``
is computed and stored in entry ``ij``.
See Also
--------
squareform : converts between condensed distance matrices and
square distance matrices.
Notes
-----
See ``squareform`` for information on how to calculate the index of
this entry or to convert the condensed distance matrix to a
redundant square matrix.
The following are common calling conventions.
1. ``Y = pdist(X, 'euclidean')``
Computes the distance between m points using Euclidean distance
(2-norm) as the distance metric between the points. The points
are arranged as m n-dimensional row vectors in the matrix X.
2. ``Y = pdist(X, 'minkowski', p=2.)``
Computes the distances using the Minkowski distance
:math:`||u-v||_p` (p-norm) where :math:`p \\geq 1`.
3. ``Y = pdist(X, 'cityblock')``
Computes the city block or Manhattan distance between the
points.
4. ``Y = pdist(X, 'seuclidean', V=None)``
Computes the standardized Euclidean distance. The standardized
Euclidean distance between two n-vectors ``u`` and ``v`` is
.. math::
\\sqrt{\\sum {(u_i-v_i)^2 / V[x_i]}}
V is the variance vector; V[i] is the variance computed over all
the i'th components of the points. If not passed, it is
automatically computed.
5. ``Y = pdist(X, 'sqeuclidean')``
Computes the squared Euclidean distance :math:`||u-v||_2^2` between
the vectors.
6. ``Y = pdist(X, 'cosine')``
Computes the cosine distance between vectors u and v,
.. math::
1 - \\frac{u \\cdot v}
{{||u||}_2 {||v||}_2}
where :math:`||*||_2` is the 2-norm of its argument ``*``, and
:math:`u \\cdot v` is the dot product of ``u`` and ``v``.
7. ``Y = pdist(X, 'correlation')``
Computes the correlation distance between vectors u and v. This is
.. math::
1 - \\frac{(u - \\bar{u}) \\cdot (v - \\bar{v})}
{{||(u - \\bar{u})||}_2 {||(v - \\bar{v})||}_2}
where :math:`\\bar{v}` is the mean of the elements of vector v,
and :math:`x \\cdot y` is the dot product of :math:`x` and :math:`y`.
8. ``Y = pdist(X, 'hamming')``
Computes the normalized Hamming distance, or the proportion of
those vector elements between two n-vectors ``u`` and ``v``
which disagree. To save memory, the matrix ``X`` can be of type
boolean.
9. ``Y = pdist(X, 'jaccard')``
Computes the Jaccard distance between the points. Given two
vectors, ``u`` and ``v``, the Jaccard distance is the
proportion of those elements ``u[i]`` and ``v[i]`` that
disagree.
10. ``Y = pdist(X, 'chebyshev')``
Computes the Chebyshev distance between the points. The
Chebyshev distance between two n-vectors ``u`` and ``v`` is the
maximum norm-1 distance between their respective elements. More
precisely, the distance is given by
.. math::
d(u,v) = \\max_i {|u_i-v_i|}
11. ``Y = pdist(X, 'canberra')``
Computes the Canberra distance between the points. The
Canberra distance between two points ``u`` and ``v`` is
.. math::
d(u,v) = \\sum_i \\frac{|u_i-v_i|}
{|u_i|+|v_i|}
12. ``Y = pdist(X, 'braycurtis')``
Computes the Bray-Curtis distance between the points. The
Bray-Curtis distance between two points ``u`` and ``v`` is
.. math::
d(u,v) = \\frac{\\sum_i {|u_i-v_i|}}
{\\sum_i {|u_i+v_i|}}
13. ``Y = pdist(X, 'mahalanobis', VI=None)``
Computes the Mahalanobis distance between the points. The
Mahalanobis distance between two points ``u`` and ``v`` is
:math:`\\sqrt{(u-v)(1/V)(u-v)^T}` where :math:`(1/V)` (the ``VI``
variable) is the inverse covariance. If ``VI`` is not None,
``VI`` will be used as the inverse covariance matrix.
14. ``Y = pdist(X, 'yule')``
Computes the Yule distance between each pair of boolean
vectors. (see yule function documentation)
15. ``Y = pdist(X, 'matching')``
Synonym for 'hamming'.
16. ``Y = pdist(X, 'dice')``
Computes the Dice distance between each pair of boolean
vectors. (see dice function documentation)
17. ``Y = pdist(X, 'kulsinski')``
Computes the Kulsinski distance between each pair of
boolean vectors. (see kulsinski function documentation)
18. ``Y = pdist(X, 'rogerstanimoto')``
Computes the Rogers-Tanimoto distance between each pair of
boolean vectors. (see rogerstanimoto function documentation)
19. ``Y = pdist(X, 'russellrao')``
Computes the Russell-Rao distance between each pair of
boolean vectors. (see russellrao function documentation)
20. ``Y = pdist(X, 'sokalmichener')``
Computes the Sokal-Michener distance between each pair of
boolean vectors. (see sokalmichener function documentation)
21. ``Y = pdist(X, 'sokalsneath')``
Computes the Sokal-Sneath distance between each pair of
boolean vectors. (see sokalsneath function documentation)
22. ``Y = pdist(X, 'wminkowski', p=2, w=w)``
Computes the weighted Minkowski distance between each pair of
vectors. (see wminkowski function documentation)
23. ``Y = pdist(X, f)``
Computes the distance between all pairs of vectors in X
using the user supplied 2-arity function f. For example,
Euclidean distance between the vectors could be computed
as follows::
dm = pdist(X, lambda u, v: np.sqrt(((u-v)**2).sum()))
Note that you should avoid passing a reference to one of
the distance functions defined in this library. For example,::
dm = pdist(X, sokalsneath)
would calculate the pair-wise distances between the vectors in
X using the Python function sokalsneath. This would result in
sokalsneath being called :math:`{n \\choose 2}` times, which
is inefficient. Instead, the optimized C version is more
efficient, and we call it using the following syntax.::
dm = pdist(X, 'sokalsneath')
"""
# You can also call this as:
# Y = pdist(X, 'test_abc')
# where 'abc' is the metric being tested. This computes the distance
# between all pairs of vectors in X using the distance metric 'abc' but
# with a more succinct, verifiable, but less efficient implementation.
X = _asarray_validated(X, sparse_ok=False, objects_ok=True, mask_ok=True,
check_finite=False)
kwargs = _args_to_kwargs_xdist(args, kwargs, metric, "pdist")
X = np.asarray(X, order='c')
s = X.shape
if len(s) != 2:
raise ValueError('A 2-dimensional array must be passed.')
m, n = s
out = kwargs.pop("out", None)
if out is None:
dm = np.empty((m * (m - 1)) // 2, dtype=np.double)
else:
if out.shape != (m * (m - 1) // 2,):
raise ValueError("output array has incorrect shape.")
if not out.flags.c_contiguous:
raise ValueError("Output array must be C-contiguous.")
if out.dtype != np.double:
raise ValueError("Output array must be double type.")
dm = out
# compute blacklist for deprecated kwargs
if(metric in _METRICS['jensenshannon'].aka
or metric == 'test_jensenshannon' or metric == jensenshannon):
kwargs_blacklist = ["p", "w", "V", "VI"]
elif(metric in _METRICS['minkowski'].aka
or metric in _METRICS['wminkowski'].aka
or metric in ['test_minkowski', 'test_wminkowski']
or metric in [minkowski, wminkowski]):
kwargs_blacklist = ["V", "VI"]
elif(metric in _METRICS['seuclidean'].aka or
metric == 'test_seuclidean' or metric == seuclidean):
kwargs_blacklist = ["p", "w", "VI"]
elif(metric in _METRICS['mahalanobis'].aka
or metric == 'test_mahalanobis' or metric == mahalanobis):
kwargs_blacklist = ["p", "w", "V"]
else:
kwargs_blacklist = ["p", "V", "VI"]
_filter_deprecated_kwargs(kwargs, kwargs_blacklist)
if callable(metric):
mstr = getattr(metric, '__name__', 'UnknownCustomMetric')
metric_name = _METRIC_ALIAS.get(mstr, None)
if metric_name is not None:
X, typ, kwargs = _validate_pdist_input(X, m, n,
metric_name, **kwargs)
k = 0
for i in xrange(0, m - 1):
for j in xrange(i + 1, m):
dm[k] = metric(X[i], X[j], **kwargs)
k = k + 1
elif isinstance(metric, string_types):
mstr = metric.lower()
mstr, kwargs = _select_weighted_metric(mstr, kwargs, out)
metric_name = _METRIC_ALIAS.get(mstr, None)
if metric_name is not None:
X, typ, kwargs = _validate_pdist_input(X, m, n,
metric_name, **kwargs)
# get pdist wrapper
pdist_fn = getattr(_distance_wrap,
"pdist_%s_%s_wrap" % (metric_name, typ))
pdist_fn(X, dm, **kwargs)
return dm
elif mstr in ['old_cosine', 'old_cos']:
warnings.warn('"old_cosine" is deprecated and will be removed in '
'a future version. Use "cosine" instead.',
DeprecationWarning)
X = _convert_to_double(X)
norms = np.einsum('ij,ij->i', X, X, dtype=np.double)
np.sqrt(norms, out=norms)
nV = norms.reshape(m, 1)
# The numerator u * v
nm = np.dot(X, X.T)
# The denom. ||u||*||v||
de = np.dot(nV, nV.T)
dm = 1.0 - (nm / de)
dm[xrange(0, m), xrange(0, m)] = 0.0
dm = squareform(dm)
elif mstr.startswith("test_"):
if mstr in _TEST_METRICS:
dm = pdist(X, _TEST_METRICS[mstr], **kwargs)
else:
raise ValueError('Unknown "Test" Distance Metric: %s' % mstr[5:])
else:
raise ValueError('Unknown Distance Metric: %s' % mstr)
else:
raise TypeError('2nd argument metric must be a string identifier '
'or a function.')
return dm
def squareform(X, force="no", checks=True):
"""
Convert a vector-form distance vector to a square-form distance
matrix, and vice-versa.
Parameters
----------
X : ndarray
Either a condensed or redundant distance matrix.
force : str, optional
As with MATLAB(TM), if force is equal to ``'tovector'`` or
``'tomatrix'``, the input will be treated as a distance matrix or
distance vector respectively.
checks : bool, optional
If set to False, no checks will be made for matrix
symmetry nor zero diagonals. This is useful if it is known that
``X - X.T1`` is small and ``diag(X)`` is close to zero.
These values are ignored any way so they do not disrupt the
squareform transformation.
Returns
-------
Y : ndarray
If a condensed distance matrix is passed, a redundant one is
returned, or if a redundant one is passed, a condensed distance
matrix is returned.
Notes
-----
1. v = squareform(X)
Given a square d-by-d symmetric distance matrix X,
``v = squareform(X)`` returns a ``d * (d-1) / 2`` (or
:math:`{n \\choose 2}`) sized vector v.
:math:`v[{n \\choose 2}-{n-i \\choose 2} + (j-i-1)]` is the distance
between points i and j. If X is non-square or asymmetric, an error
is returned.
2. X = squareform(v)
Given a ``d*(d-1)/2`` sized v for some integer ``d >= 2`` encoding
distances as described, ``X = squareform(v)`` returns a d by d distance
matrix X. The ``X[i, j]`` and ``X[j, i]`` values are set to
:math:`v[{n \\choose 2}-{n-i \\choose 2} + (j-i-1)]` and all
diagonal elements are zero.
In SciPy 0.19.0, ``squareform`` stopped casting all input types to
float64, and started returning arrays of the same dtype as the input.
"""
X = np.ascontiguousarray(X)
s = X.shape
if force.lower() == 'tomatrix':
if len(s) != 1:
raise ValueError("Forcing 'tomatrix' but input X is not a "
"distance vector.")
elif force.lower() == 'tovector':
if len(s) != 2:
raise ValueError("Forcing 'tovector' but input X is not a "
"distance matrix.")
# X = squareform(v)
if len(s) == 1:
if s[0] == 0:
return np.zeros((1, 1), dtype=X.dtype)
# Grab the closest value to the square root of the number
# of elements times 2 to see if the number of elements
# is indeed a binomial coefficient.
d = int(np.ceil(np.sqrt(s[0] * 2)))
# Check that v is of valid dimensions.
if d * (d - 1) != s[0] * 2:
raise ValueError('Incompatible vector size. It must be a binomial '
'coefficient n choose 2 for some integer n >= 2.')
# Allocate memory for the distance matrix.
M = np.zeros((d, d), dtype=X.dtype)
# Since the C code does not support striding using strides.
# The dimensions are used instead.
X = _copy_array_if_base_present(X)
# Fill in the values of the distance matrix.
_distance_wrap.to_squareform_from_vector_wrap(M, X)
# Return the distance matrix.
return M
elif len(s) == 2:
if s[0] != s[1]:
raise ValueError('The matrix argument must be square.')
if checks:
is_valid_dm(X, throw=True, name='X')
# One-side of the dimensions is set here.
d = s[0]
if d <= 1:
return np.array([], dtype=X.dtype)
# Create a vector.
v = np.zeros((d * (d - 1)) // 2, dtype=X.dtype)
# Since the C code does not support striding using strides.
# The dimensions are used instead.
X = _copy_array_if_base_present(X)
# Convert the vector to squareform.
_distance_wrap.to_vector_from_squareform_wrap(X, v)
return v
else:
raise ValueError(('The first argument must be one or two dimensional '
'array. A %d-dimensional array is not '
'permitted') % len(s))
def is_valid_dm(D, tol=0.0, throw=False, name="D", warning=False):
"""
Return True if input array is a valid distance matrix.
Distance matrices must be 2-dimensional numpy arrays.
They must have a zero-diagonal, and they must be symmetric.
Parameters
----------
D : ndarray
The candidate object to test for validity.
tol : float, optional
The distance matrix should be symmetric. `tol` is the maximum
difference between entries ``ij`` and ``ji`` for the distance
metric to be considered symmetric.
throw : bool, optional
An exception is thrown if the distance matrix passed is not valid.
name : str, optional
The name of the variable to checked. This is useful if
throw is set to True so the offending variable can be identified
in the exception message when an exception is thrown.
warning : bool, optional
Instead of throwing an exception, a warning message is
raised.
Returns
-------
valid : bool
True if the variable `D` passed is a valid distance matrix.
Notes
-----
Small numerical differences in `D` and `D.T` and non-zeroness of
the diagonal are ignored if they are within the tolerance specified
by `tol`.
"""
D = np.asarray(D, order='c')
valid = True
try:
s = D.shape
if len(D.shape) != 2:
if name:
raise ValueError(('Distance matrix \'%s\' must have shape=2 '
'(i.e. be two-dimensional).') % name)
else:
raise ValueError('Distance matrix must have shape=2 (i.e. '
'be two-dimensional).')
if tol == 0.0:
if not (D == D.T).all():
if name:
raise ValueError(('Distance matrix \'%s\' must be '
'symmetric.') % name)
else:
raise ValueError('Distance matrix must be symmetric.')
if not (D[xrange(0, s[0]), xrange(0, s[0])] == 0).all():
if name:
raise ValueError(('Distance matrix \'%s\' diagonal must '
'be zero.') % name)
else:
raise ValueError('Distance matrix diagonal must be zero.')
else:
if not (D - D.T <= tol).all():
if name:
raise ValueError(('Distance matrix \'%s\' must be '
'symmetric within tolerance %5.5f.')
% (name, tol))
else:
raise ValueError('Distance matrix must be symmetric within'
' tolerance %5.5f.' % tol)
if not (D[xrange(0, s[0]), xrange(0, s[0])] <= tol).all():
if name:
raise ValueError(('Distance matrix \'%s\' diagonal must be'
' close to zero within tolerance %5.5f.')
% (name, tol))
else:
raise ValueError(('Distance matrix \'%s\' diagonal must be'
' close to zero within tolerance %5.5f.')
% tol)
except Exception as e:
if throw:
raise
if warning:
warnings.warn(str(e))
valid = False
return valid
def is_valid_y(y, warning=False, throw=False, name=None):
"""
Return True if the input array is a valid condensed distance matrix.
Condensed distance matrices must be 1-dimensional numpy arrays.
Their length must be a binomial coefficient :math:`{n \\choose 2}`
for some positive integer n.
Parameters
----------
y : ndarray
The condensed distance matrix.
warning : bool, optional
Invokes a warning if the variable passed is not a valid
condensed distance matrix. The warning message explains why
the distance matrix is not valid. `name` is used when
referencing the offending variable.
throw : bool, optional
Throws an exception if the variable passed is not a valid
condensed distance matrix.
name : bool, optional
Used when referencing the offending variable in the
warning or exception message.
"""
y = np.asarray(y, order='c')
valid = True
try:
if len(y.shape) != 1:
if name:
raise ValueError(('Condensed distance matrix \'%s\' must '
'have shape=1 (i.e. be one-dimensional).')
% name)
else:
raise ValueError('Condensed distance matrix must have shape=1 '
'(i.e. be one-dimensional).')
n = y.shape[0]
d = int(np.ceil(np.sqrt(n * 2)))
if (d * (d - 1) / 2) != n:
if name:
raise ValueError(('Length n of condensed distance matrix '
'\'%s\' must be a binomial coefficient, i.e.'
'there must be a k such that '
'(k \\choose 2)=n)!') % name)
else:
raise ValueError('Length n of condensed distance matrix must '
'be a binomial coefficient, i.e. there must '
'be a k such that (k \\choose 2)=n)!')
except Exception as e:
if throw:
raise
if warning:
warnings.warn(str(e))
valid = False
return valid
def num_obs_dm(d):
"""
Return the number of original observations that correspond to a
square, redundant distance matrix.
Parameters
----------
d : ndarray
The target distance matrix.
Returns
-------
num_obs_dm : int
The number of observations in the redundant distance matrix.
"""
d = np.asarray(d, order='c')
is_valid_dm(d, tol=np.inf, throw=True, name='d')
return d.shape[0]
def num_obs_y(Y):
"""
Return the number of original observations that correspond to a
condensed distance matrix.
Parameters
----------
Y : ndarray
Condensed distance matrix.
Returns
-------
n : int
The number of observations in the condensed distance matrix `Y`.
"""
Y = np.asarray(Y, order='c')
is_valid_y(Y, throw=True, name='Y')
k = Y.shape[0]
if k == 0:
raise ValueError("The number of observations cannot be determined on "
"an empty distance matrix.")
d = int(np.ceil(np.sqrt(k * 2)))
if (d * (d - 1) / 2) != k:
raise ValueError("Invalid condensed distance matrix passed. Must be "
"some k where k=(n choose 2) for some n >= 2.")
return d
def cdist(XA, XB, metric='euclidean', *args, **kwargs):
"""
Compute distance between each pair of the two collections of inputs.
See Notes for common calling conventions.
Parameters
----------
XA : ndarray
An :math:`m_A` by :math:`n` array of :math:`m_A`
original observations in an :math:`n`-dimensional space.
Inputs are converted to float type.
XB : ndarray
An :math:`m_B` by :math:`n` array of :math:`m_B`
original observations in an :math:`n`-dimensional space.
Inputs are converted to float type.
metric : str or callable, optional
The distance metric to use. If a string, the distance function can be
'braycurtis', 'canberra', 'chebyshev', 'cityblock', 'correlation',
'cosine', 'dice', 'euclidean', 'hamming', 'jaccard', 'jensenshannon',
'kulsinski', 'mahalanobis', 'matching', 'minkowski', 'rogerstanimoto',
'russellrao', 'seuclidean', 'sokalmichener', 'sokalsneath', 'sqeuclidean',
'wminkowski', 'yule'.
*args : tuple. Deprecated.
Additional arguments should be passed as keyword arguments
**kwargs : dict, optional
Extra arguments to `metric`: refer to each metric documentation for a
list of all possible arguments.
Some possible arguments:
p : scalar
The p-norm to apply for Minkowski, weighted and unweighted.
Default: 2.
w : ndarray
The weight vector for metrics that support weights (e.g., Minkowski).
V : ndarray
The variance vector for standardized Euclidean.
Default: var(vstack([XA, XB]), axis=0, ddof=1)
VI : ndarray
The inverse of the covariance matrix for Mahalanobis.
Default: inv(cov(vstack([XA, XB].T))).T
out : ndarray
The output array
If not None, the distance matrix Y is stored in this array.
Note: metric independent, it will become a regular keyword arg in a
future scipy version
Returns
-------
Y : ndarray
A :math:`m_A` by :math:`m_B` distance matrix is returned.
For each :math:`i` and :math:`j`, the metric
``dist(u=XA[i], v=XB[j])`` is computed and stored in the
:math:`ij` th entry.
Raises
------
ValueError
An exception is thrown if `XA` and `XB` do not have
the same number of columns.
Notes
-----
The following are common calling conventions:
1. ``Y = cdist(XA, XB, 'euclidean')``
Computes the distance between :math:`m` points using
Euclidean distance (2-norm) as the distance metric between the
points. The points are arranged as :math:`m`
:math:`n`-dimensional row vectors in the matrix X.
2. ``Y = cdist(XA, XB, 'minkowski', p=2.)``
Computes the distances using the Minkowski distance
:math:`||u-v||_p` (:math:`p`-norm) where :math:`p \\geq 1`.
3. ``Y = cdist(XA, XB, 'cityblock')``
Computes the city block or Manhattan distance between the
points.
4. ``Y = cdist(XA, XB, 'seuclidean', V=None)``
Computes the standardized Euclidean distance. The standardized
Euclidean distance between two n-vectors ``u`` and ``v`` is
.. math::
\\sqrt{\\sum {(u_i-v_i)^2 / V[x_i]}}.
V is the variance vector; V[i] is the variance computed over all
the i'th components of the points. If not passed, it is
automatically computed.
5. ``Y = cdist(XA, XB, 'sqeuclidean')``
Computes the squared Euclidean distance :math:`||u-v||_2^2` between
the vectors.
6. ``Y = cdist(XA, XB, 'cosine')``
Computes the cosine distance between vectors u and v,
.. math::
1 - \\frac{u \\cdot v}
{{||u||}_2 {||v||}_2}
where :math:`||*||_2` is the 2-norm of its argument ``*``, and
:math:`u \\cdot v` is the dot product of :math:`u` and :math:`v`.
7. ``Y = cdist(XA, XB, 'correlation')``
Computes the correlation distance between vectors u and v. This is
.. math::
1 - \\frac{(u - \\bar{u}) \\cdot (v - \\bar{v})}
{{||(u - \\bar{u})||}_2 {||(v - \\bar{v})||}_2}
where :math:`\\bar{v}` is the mean of the elements of vector v,
and :math:`x \\cdot y` is the dot product of :math:`x` and :math:`y`.
8. ``Y = cdist(XA, XB, 'hamming')``
Computes the normalized Hamming distance, or the proportion of
those vector elements between two n-vectors ``u`` and ``v``
which disagree. To save memory, the matrix ``X`` can be of type
boolean.
9. ``Y = cdist(XA, XB, 'jaccard')``
Computes the Jaccard distance between the points. Given two
vectors, ``u`` and ``v``, the Jaccard distance is the
proportion of those elements ``u[i]`` and ``v[i]`` that
disagree where at least one of them is non-zero.
10. ``Y = cdist(XA, XB, 'chebyshev')``
Computes the Chebyshev distance between the points. The
Chebyshev distance between two n-vectors ``u`` and ``v`` is the
maximum norm-1 distance between their respective elements. More
precisely, the distance is given by
.. math::
d(u,v) = \\max_i {|u_i-v_i|}.
11. ``Y = cdist(XA, XB, 'canberra')``
Computes the Canberra distance between the points. The
Canberra distance between two points ``u`` and ``v`` is
.. math::
d(u,v) = \\sum_i \\frac{|u_i-v_i|}
{|u_i|+|v_i|}.
12. ``Y = cdist(XA, XB, 'braycurtis')``
Computes the Bray-Curtis distance between the points. The
Bray-Curtis distance between two points ``u`` and ``v`` is
.. math::
d(u,v) = \\frac{\\sum_i (|u_i-v_i|)}
{\\sum_i (|u_i+v_i|)}
13. ``Y = cdist(XA, XB, 'mahalanobis', VI=None)``
Computes the Mahalanobis distance between the points. The
Mahalanobis distance between two points ``u`` and ``v`` is
:math:`\\sqrt{(u-v)(1/V)(u-v)^T}` where :math:`(1/V)` (the ``VI``
variable) is the inverse covariance. If ``VI`` is not None,
``VI`` will be used as the inverse covariance matrix.
14. ``Y = cdist(XA, XB, 'yule')``
Computes the Yule distance between the boolean
vectors. (see `yule` function documentation)
15. ``Y = cdist(XA, XB, 'matching')``
Synonym for 'hamming'.
16. ``Y = cdist(XA, XB, 'dice')``
Computes the Dice distance between the boolean vectors. (see
`dice` function documentation)
17. ``Y = cdist(XA, XB, 'kulsinski')``
Computes the Kulsinski distance between the boolean
vectors. (see `kulsinski` function documentation)
18. ``Y = cdist(XA, XB, 'rogerstanimoto')``
Computes the Rogers-Tanimoto distance between the boolean
vectors. (see `rogerstanimoto` function documentation)
19. ``Y = cdist(XA, XB, 'russellrao')``
Computes the Russell-Rao distance between the boolean
vectors. (see `russellrao` function documentation)
20. ``Y = cdist(XA, XB, 'sokalmichener')``
Computes the Sokal-Michener distance between the boolean
vectors. (see `sokalmichener` function documentation)
21. ``Y = cdist(XA, XB, 'sokalsneath')``
Computes the Sokal-Sneath distance between the vectors. (see
`sokalsneath` function documentation)
22. ``Y = cdist(XA, XB, 'wminkowski', p=2., w=w)``
Computes the weighted Minkowski distance between the
vectors. (see `wminkowski` function documentation)
23. ``Y = cdist(XA, XB, f)``
Computes the distance between all pairs of vectors in X
using the user supplied 2-arity function f. For example,
Euclidean distance between the vectors could be computed
as follows::
dm = cdist(XA, XB, lambda u, v: np.sqrt(((u-v)**2).sum()))
Note that you should avoid passing a reference to one of
the distance functions defined in this library. For example,::
dm = cdist(XA, XB, sokalsneath)
would calculate the pair-wise distances between the vectors in
X using the Python function `sokalsneath`. This would result in
sokalsneath being called :math:`{n \\choose 2}` times, which
is inefficient. Instead, the optimized C version is more
efficient, and we call it using the following syntax::
dm = cdist(XA, XB, 'sokalsneath')
Examples
--------
Find the Euclidean distances between four 2-D coordinates:
>>> from scipy.spatial import distance
>>> coords = [(35.0456, -85.2672),
... (35.1174, -89.9711),
... (35.9728, -83.9422),
... (36.1667, -86.7833)]
>>> distance.cdist(coords, coords, 'euclidean')
array([[ 0. , 4.7044, 1.6172, 1.8856],
[ 4.7044, 0. , 6.0893, 3.3561],
[ 1.6172, 6.0893, 0. , 2.8477],
[ 1.8856, 3.3561, 2.8477, 0. ]])
Find the Manhattan distance from a 3-D point to the corners of the unit
cube:
>>> a = np.array([[0, 0, 0],
... [0, 0, 1],
... [0, 1, 0],
... [0, 1, 1],
... [1, 0, 0],
... [1, 0, 1],
... [1, 1, 0],
... [1, 1, 1]])
>>> b = np.array([[ 0.1, 0.2, 0.4]])
>>> distance.cdist(a, b, 'cityblock')
array([[ 0.7],
[ 0.9],
[ 1.3],
[ 1.5],
[ 1.5],
[ 1.7],
[ 2.1],
[ 2.3]])
"""
# You can also call this as:
# Y = cdist(XA, XB, 'test_abc')
# where 'abc' is the metric being tested. This computes the distance
# between all pairs of vectors in XA and XB using the distance metric 'abc'
# but with a more succinct, verifiable, but less efficient implementation.
kwargs = _args_to_kwargs_xdist(args, kwargs, metric, "cdist")
XA = np.asarray(XA, order='c')
XB = np.asarray(XB, order='c')
s = XA.shape
sB = XB.shape
if len(s) != 2:
raise ValueError('XA must be a 2-dimensional array.')
if len(sB) != 2:
raise ValueError('XB must be a 2-dimensional array.')
if s[1] != sB[1]:
raise ValueError('XA and XB must have the same number of columns '
'(i.e. feature dimension.)')
mA = s[0]
mB = sB[0]
n = s[1]
out = kwargs.pop("out", None)
if out is None:
dm = np.empty((mA, mB), dtype=np.double)
else:
if out.shape != (mA, mB):
raise ValueError("Output array has incorrect shape.")
if not out.flags.c_contiguous:
raise ValueError("Output array must be C-contiguous.")
if out.dtype != np.double:
raise ValueError("Output array must be double type.")
dm = out
# compute blacklist for deprecated kwargs
if(metric in _METRICS['minkowski'].aka or
metric in _METRICS['wminkowski'].aka or
metric in ['test_minkowski', 'test_wminkowski'] or
metric in [minkowski, wminkowski]):
kwargs_blacklist = ["V", "VI"]
elif(metric in _METRICS['seuclidean'].aka or
metric == 'test_seuclidean' or metric == seuclidean):
kwargs_blacklist = ["p", "w", "VI"]
elif(metric in _METRICS['mahalanobis'].aka or
metric == 'test_mahalanobis' or metric == mahalanobis):
kwargs_blacklist = ["p", "w", "V"]
else:
kwargs_blacklist = ["p", "V", "VI"]
_filter_deprecated_kwargs(kwargs, kwargs_blacklist)
if callable(metric):
mstr = getattr(metric, '__name__', 'Unknown')
metric_name = _METRIC_ALIAS.get(mstr, None)
XA, XB, typ, kwargs = _validate_cdist_input(XA, XB, mA, mB, n,
metric_name, **kwargs)
for i in xrange(0, mA):
for j in xrange(0, mB):
dm[i, j] = metric(XA[i], XB[j], **kwargs)
elif isinstance(metric, string_types):
mstr = metric.lower()
mstr, kwargs = _select_weighted_metric(mstr, kwargs, out)
metric_name = _METRIC_ALIAS.get(mstr, None)
if metric_name is not None:
XA, XB, typ, kwargs = _validate_cdist_input(XA, XB, mA, mB, n,
metric_name, **kwargs)
# get cdist wrapper
cdist_fn = getattr(_distance_wrap,
"cdist_%s_%s_wrap" % (metric_name, typ))
cdist_fn(XA, XB, dm, **kwargs)
return dm
elif mstr.startswith("test_"):
if mstr in _TEST_METRICS:
dm = cdist(XA, XB, _TEST_METRICS[mstr], **kwargs)
else:
raise ValueError('Unknown "Test" Distance Metric: %s' % mstr[5:])
else:
raise ValueError('Unknown Distance Metric: %s' % mstr)
else:
raise TypeError('2nd argument metric must be a string identifier '
'or a function.')
return dm