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
|
/*
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to You under the Apache License, Version 2.0
* (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.spark.mllib.api.python
import java.io.OutputStream
import java.nio.{ByteBuffer, ByteOrder}
import java.util.{ArrayList => JArrayList, List => JList, Map => JMap}
import scala.collection.JavaConverters._
import scala.collection.mutable.ArrayBuffer
import scala.language.existentials
import scala.reflect.ClassTag
import net.razorvine.pickle._
import org.apache.spark.SparkContext
import org.apache.spark.api.java.{JavaRDD, JavaSparkContext}
import org.apache.spark.api.python.SerDeUtil
import org.apache.spark.mllib.classification._
import org.apache.spark.mllib.clustering._
import org.apache.spark.mllib.evaluation.RankingMetrics
import org.apache.spark.mllib.feature._
import org.apache.spark.mllib.fpm.{FPGrowth, FPGrowthModel, PrefixSpan}
import org.apache.spark.mllib.linalg._
import org.apache.spark.mllib.linalg.distributed._
import org.apache.spark.mllib.optimization._
import org.apache.spark.mllib.random.{RandomRDDs => RG}
import org.apache.spark.mllib.recommendation._
import org.apache.spark.mllib.regression._
import org.apache.spark.mllib.stat.correlation.CorrelationNames
import org.apache.spark.mllib.stat.distribution.MultivariateGaussian
import org.apache.spark.mllib.stat.test.{ChiSqTestResult, KolmogorovSmirnovTestResult}
import org.apache.spark.mllib.stat.{
KernelDensity, MultivariateStatisticalSummary, Statistics}
import org.apache.spark.mllib.tree.configuration.{Algo, BoostingStrategy, Strategy}
import org.apache.spark.mllib.tree.impurity._
import org.apache.spark.mllib.tree.loss.Losses
import org.apache.spark.mllib.tree.model.{DecisionTreeModel, GradientBoostedTreesModel, RandomForestModel}
import org.apache.spark.mllib.tree.{DecisionTree, GradientBoostedTrees, RandomForest}
import org.apache.spark.mllib.util.MLUtils
import org.apache.spark.mllib.util.LinearDataGenerator
import org.apache.spark.rdd.RDD
import org.apache.spark.sql.{DataFrame, Row, SQLContext}
import org.apache.spark.storage.StorageLevel
import org.apache.spark.util.Utils
/**
* The Java stubs necessary for the Python mllib bindings. It is called by Py4J on the Python side.
*/
private[python] class PythonMLLibAPI extends Serializable {
/**
* Loads and serializes labeled points saved with `RDD#saveAsTextFile`.
* @param jsc Java SparkContext
* @param path file or directory path in any Hadoop-supported file system URI
* @param minPartitions min number of partitions
* @return serialized labeled points stored in a JavaRDD of byte array
*/
def loadLabeledPoints(
jsc: JavaSparkContext,
path: String,
minPartitions: Int): JavaRDD[LabeledPoint] =
MLUtils.loadLabeledPoints(jsc.sc, path, minPartitions)
/**
* Loads and serializes vectors saved with `RDD#saveAsTextFile`.
* @param jsc Java SparkContext
* @param path file or directory path in any Hadoop-supported file system URI
* @return serialized vectors in a RDD
*/
def loadVectors(jsc: JavaSparkContext, path: String): RDD[Vector] =
MLUtils.loadVectors(jsc.sc, path)
private def trainRegressionModel(
learner: GeneralizedLinearAlgorithm[_ <: GeneralizedLinearModel],
data: JavaRDD[LabeledPoint],
initialWeights: Vector): JList[Object] = {
try {
val model = learner.run(data.rdd.persist(StorageLevel.MEMORY_AND_DISK), initialWeights)
if (model.isInstanceOf[LogisticRegressionModel]) {
val lrModel = model.asInstanceOf[LogisticRegressionModel]
List(lrModel.weights, lrModel.intercept, lrModel.numFeatures, lrModel.numClasses)
.map(_.asInstanceOf[Object]).asJava
} else {
List(model.weights, model.intercept).map(_.asInstanceOf[Object]).asJava
}
} finally {
data.rdd.unpersist(blocking = false)
}
}
/**
* Return the Updater from string
*/
def getUpdaterFromString(regType: String): Updater = {
if (regType == "l2") {
new SquaredL2Updater
} else if (regType == "l1") {
new L1Updater
} else if (regType == null || regType == "none") {
new SimpleUpdater
} else {
throw new IllegalArgumentException("Invalid value for 'regType' parameter."
+ " Can only be initialized using the following string values: ['l1', 'l2', None].")
}
}
/**
* Java stub for Python mllib LinearRegressionWithSGD.train()
*/
def trainLinearRegressionModelWithSGD(
data: JavaRDD[LabeledPoint],
numIterations: Int,
stepSize: Double,
miniBatchFraction: Double,
initialWeights: Vector,
regParam: Double,
regType: String,
intercept: Boolean,
validateData: Boolean,
convergenceTol: Double): JList[Object] = {
val lrAlg = new LinearRegressionWithSGD()
lrAlg.setIntercept(intercept)
.setValidateData(validateData)
lrAlg.optimizer
.setNumIterations(numIterations)
.setRegParam(regParam)
.setStepSize(stepSize)
.setMiniBatchFraction(miniBatchFraction)
.setConvergenceTol(convergenceTol)
lrAlg.optimizer.setUpdater(getUpdaterFromString(regType))
trainRegressionModel(
lrAlg,
data,
initialWeights)
}
/**
* Java stub for Python mllib LassoWithSGD.train()
*/
def trainLassoModelWithSGD(
data: JavaRDD[LabeledPoint],
numIterations: Int,
stepSize: Double,
regParam: Double,
miniBatchFraction: Double,
initialWeights: Vector,
intercept: Boolean,
validateData: Boolean,
convergenceTol: Double): JList[Object] = {
val lassoAlg = new LassoWithSGD()
lassoAlg.setIntercept(intercept)
.setValidateData(validateData)
lassoAlg.optimizer
.setNumIterations(numIterations)
.setRegParam(regParam)
.setStepSize(stepSize)
.setMiniBatchFraction(miniBatchFraction)
.setConvergenceTol(convergenceTol)
trainRegressionModel(
lassoAlg,
data,
initialWeights)
}
/**
* Java stub for Python mllib RidgeRegressionWithSGD.train()
*/
def trainRidgeModelWithSGD(
data: JavaRDD[LabeledPoint],
numIterations: Int,
stepSize: Double,
regParam: Double,
miniBatchFraction: Double,
initialWeights: Vector,
intercept: Boolean,
validateData: Boolean,
convergenceTol: Double): JList[Object] = {
val ridgeAlg = new RidgeRegressionWithSGD()
ridgeAlg.setIntercept(intercept)
.setValidateData(validateData)
ridgeAlg.optimizer
.setNumIterations(numIterations)
.setRegParam(regParam)
.setStepSize(stepSize)
.setMiniBatchFraction(miniBatchFraction)
.setConvergenceTol(convergenceTol)
trainRegressionModel(
ridgeAlg,
data,
initialWeights)
}
/**
* Java stub for Python mllib SVMWithSGD.train()
*/
def trainSVMModelWithSGD(
data: JavaRDD[LabeledPoint],
numIterations: Int,
stepSize: Double,
regParam: Double,
miniBatchFraction: Double,
initialWeights: Vector,
regType: String,
intercept: Boolean,
validateData: Boolean,
convergenceTol: Double): JList[Object] = {
val SVMAlg = new SVMWithSGD()
SVMAlg.setIntercept(intercept)
.setValidateData(validateData)
SVMAlg.optimizer
.setNumIterations(numIterations)
.setRegParam(regParam)
.setStepSize(stepSize)
.setMiniBatchFraction(miniBatchFraction)
.setConvergenceTol(convergenceTol)
SVMAlg.optimizer.setUpdater(getUpdaterFromString(regType))
trainRegressionModel(
SVMAlg,
data,
initialWeights)
}
/**
* Java stub for Python mllib LogisticRegressionWithSGD.train()
*/
def trainLogisticRegressionModelWithSGD(
data: JavaRDD[LabeledPoint],
numIterations: Int,
stepSize: Double,
miniBatchFraction: Double,
initialWeights: Vector,
regParam: Double,
regType: String,
intercept: Boolean,
validateData: Boolean,
convergenceTol: Double): JList[Object] = {
val LogRegAlg = new LogisticRegressionWithSGD()
LogRegAlg.setIntercept(intercept)
.setValidateData(validateData)
LogRegAlg.optimizer
.setNumIterations(numIterations)
.setRegParam(regParam)
.setStepSize(stepSize)
.setMiniBatchFraction(miniBatchFraction)
.setConvergenceTol(convergenceTol)
LogRegAlg.optimizer.setUpdater(getUpdaterFromString(regType))
trainRegressionModel(
LogRegAlg,
data,
initialWeights)
}
/**
* Java stub for Python mllib LogisticRegressionWithLBFGS.train()
*/
def trainLogisticRegressionModelWithLBFGS(
data: JavaRDD[LabeledPoint],
numIterations: Int,
initialWeights: Vector,
regParam: Double,
regType: String,
intercept: Boolean,
corrections: Int,
tolerance: Double,
validateData: Boolean,
numClasses: Int): JList[Object] = {
val LogRegAlg = new LogisticRegressionWithLBFGS()
LogRegAlg.setIntercept(intercept)
.setValidateData(validateData)
.setNumClasses(numClasses)
LogRegAlg.optimizer
.setNumIterations(numIterations)
.setRegParam(regParam)
.setNumCorrections(corrections)
.setConvergenceTol(tolerance)
LogRegAlg.optimizer.setUpdater(getUpdaterFromString(regType))
trainRegressionModel(
LogRegAlg,
data,
initialWeights)
}
/**
* Java stub for NaiveBayes.train()
*/
def trainNaiveBayesModel(
data: JavaRDD[LabeledPoint],
lambda: Double): JList[Object] = {
val model = NaiveBayes.train(data.rdd, lambda)
List(Vectors.dense(model.labels), Vectors.dense(model.pi), model.theta.map(Vectors.dense)).
map(_.asInstanceOf[Object]).asJava
}
/**
* Java stub for Python mllib IsotonicRegression.run()
*/
def trainIsotonicRegressionModel(
data: JavaRDD[Vector],
isotonic: Boolean): JList[Object] = {
val isotonicRegressionAlg = new IsotonicRegression().setIsotonic(isotonic)
val input = data.rdd.map { x =>
(x(0), x(1), x(2))
}.persist(StorageLevel.MEMORY_AND_DISK)
try {
val model = isotonicRegressionAlg.run(input)
List[AnyRef](model.boundaryVector, model.predictionVector).asJava
} finally {
data.rdd.unpersist(blocking = false)
}
}
/**
* Java stub for Python mllib KMeans.run()
*/
def trainKMeansModel(
data: JavaRDD[Vector],
k: Int,
maxIterations: Int,
runs: Int,
initializationMode: String,
seed: java.lang.Long,
initializationSteps: Int,
epsilon: Double,
initialModel: java.util.ArrayList[Vector]): KMeansModel = {
val kMeansAlg = new KMeans()
.setK(k)
.setMaxIterations(maxIterations)
.setRuns(runs)
.setInitializationMode(initializationMode)
.setInitializationSteps(initializationSteps)
.setEpsilon(epsilon)
if (seed != null) kMeansAlg.setSeed(seed)
if (!initialModel.isEmpty()) kMeansAlg.setInitialModel(new KMeansModel(initialModel))
try {
kMeansAlg.run(data.rdd.persist(StorageLevel.MEMORY_AND_DISK))
} finally {
data.rdd.unpersist(blocking = false)
}
}
/**
* Java stub for Python mllib KMeansModel.computeCost()
*/
def computeCostKmeansModel(
data: JavaRDD[Vector],
centers: java.util.ArrayList[Vector]): Double = {
new KMeansModel(centers).computeCost(data)
}
/**
* Java stub for Python mllib GaussianMixture.run()
* Returns a list containing weights, mean and covariance of each mixture component.
*/
def trainGaussianMixtureModel(
data: JavaRDD[Vector],
k: Int,
convergenceTol: Double,
maxIterations: Int,
seed: java.lang.Long,
initialModelWeights: java.util.ArrayList[Double],
initialModelMu: java.util.ArrayList[Vector],
initialModelSigma: java.util.ArrayList[Matrix]): GaussianMixtureModelWrapper = {
val gmmAlg = new GaussianMixture()
.setK(k)
.setConvergenceTol(convergenceTol)
.setMaxIterations(maxIterations)
if (initialModelWeights != null && initialModelMu != null && initialModelSigma != null) {
val gaussians = initialModelMu.asScala.toSeq.zip(initialModelSigma.asScala.toSeq).map {
case (x, y) => new MultivariateGaussian(x.asInstanceOf[Vector], y.asInstanceOf[Matrix])
}
val initialModel = new GaussianMixtureModel(
initialModelWeights.asScala.toArray, gaussians.toArray)
gmmAlg.setInitialModel(initialModel)
}
if (seed != null) gmmAlg.setSeed(seed)
try {
new GaussianMixtureModelWrapper(gmmAlg.run(data.rdd.persist(StorageLevel.MEMORY_AND_DISK)))
} finally {
data.rdd.unpersist(blocking = false)
}
}
/**
* Java stub for Python mllib GaussianMixtureModel.predictSoft()
*/
def predictSoftGMM(
data: JavaRDD[Vector],
wt: Vector,
mu: Array[Object],
si: Array[Object]): RDD[Vector] = {
val weight = wt.toArray
val mean = mu.map(_.asInstanceOf[DenseVector])
val sigma = si.map(_.asInstanceOf[DenseMatrix])
val gaussians = Array.tabulate(weight.length){
i => new MultivariateGaussian(mean(i), sigma(i))
}
val model = new GaussianMixtureModel(weight, gaussians)
model.predictSoft(data).map(Vectors.dense)
}
/**
* Java stub for Python mllib PowerIterationClustering.run(). This stub returns a
* handle to the Java object instead of the content of the Java object. Extra care
* needs to be taken in the Python code to ensure it gets freed on exit; see the
* Py4J documentation.
* @param data an RDD of (i, j, s,,ij,,) tuples representing the affinity matrix.
* @param k number of clusters.
* @param maxIterations maximum number of iterations of the power iteration loop.
* @param initMode the initialization mode. This can be either "random" to use
* a random vector as vertex properties, or "degree" to use
* normalized sum similarities. Default: random.
*/
def trainPowerIterationClusteringModel(
data: JavaRDD[Vector],
k: Int,
maxIterations: Int,
initMode: String): PowerIterationClusteringModel = {
val pic = new PowerIterationClustering()
.setK(k)
.setMaxIterations(maxIterations)
.setInitializationMode(initMode)
val model = pic.run(data.rdd.map(v => (v(0).toLong, v(1).toLong, v(2))))
new PowerIterationClusteringModelWrapper(model)
}
/**
* Java stub for Python mllib ALS.train(). This stub returns a handle
* to the Java object instead of the content of the Java object. Extra care
* needs to be taken in the Python code to ensure it gets freed on exit; see
* the Py4J documentation.
*/
def trainALSModel(
ratingsJRDD: JavaRDD[Rating],
rank: Int,
iterations: Int,
lambda: Double,
blocks: Int,
nonnegative: Boolean,
seed: java.lang.Long): MatrixFactorizationModel = {
val als = new ALS()
.setRank(rank)
.setIterations(iterations)
.setLambda(lambda)
.setBlocks(blocks)
.setNonnegative(nonnegative)
if (seed != null) als.setSeed(seed)
val model = als.run(ratingsJRDD.rdd)
new MatrixFactorizationModelWrapper(model)
}
/**
* Java stub for Python mllib ALS.trainImplicit(). This stub returns a
* handle to the Java object instead of the content of the Java object.
* Extra care needs to be taken in the Python code to ensure it gets freed on
* exit; see the Py4J documentation.
*/
def trainImplicitALSModel(
ratingsJRDD: JavaRDD[Rating],
rank: Int,
iterations: Int,
lambda: Double,
blocks: Int,
alpha: Double,
nonnegative: Boolean,
seed: java.lang.Long): MatrixFactorizationModel = {
val als = new ALS()
.setImplicitPrefs(true)
.setRank(rank)
.setIterations(iterations)
.setLambda(lambda)
.setBlocks(blocks)
.setAlpha(alpha)
.setNonnegative(nonnegative)
if (seed != null) als.setSeed(seed)
val model = als.run(ratingsJRDD.rdd)
new MatrixFactorizationModelWrapper(model)
}
/**
* Java stub for Python mllib LDA.run()
*/
def trainLDAModel(
data: JavaRDD[java.util.List[Any]],
k: Int,
maxIterations: Int,
docConcentration: Double,
topicConcentration: Double,
seed: java.lang.Long,
checkpointInterval: Int,
optimizer: String): LDAModelWrapper = {
val algo = new LDA()
.setK(k)
.setMaxIterations(maxIterations)
.setDocConcentration(docConcentration)
.setTopicConcentration(topicConcentration)
.setCheckpointInterval(checkpointInterval)
.setOptimizer(optimizer)
if (seed != null) algo.setSeed(seed)
val documents = data.rdd.map(_.asScala.toArray).map { r =>
r(0) match {
case i: java.lang.Integer => (i.toLong, r(1).asInstanceOf[Vector])
case i: java.lang.Long => (i.toLong, r(1).asInstanceOf[Vector])
case _ => throw new IllegalArgumentException("input values contains invalid type value.")
}
}
val model = algo.run(documents)
new LDAModelWrapper(model)
}
/**
* Load a LDA model
*/
def loadLDAModel(jsc: JavaSparkContext, path: String): LDAModelWrapper = {
val model = DistributedLDAModel.load(jsc.sc, path)
new LDAModelWrapper(model)
}
/**
* Java stub for Python mllib FPGrowth.train(). This stub returns a handle
* to the Java object instead of the content of the Java object. Extra care
* needs to be taken in the Python code to ensure it gets freed on exit; see
* the Py4J documentation.
*/
def trainFPGrowthModel(
data: JavaRDD[java.lang.Iterable[Any]],
minSupport: Double,
numPartitions: Int): FPGrowthModel[Any] = {
val fpg = new FPGrowth()
.setMinSupport(minSupport)
.setNumPartitions(numPartitions)
val model = fpg.run(data.rdd.map(_.asScala.toArray))
new FPGrowthModelWrapper(model)
}
/**
* Java stub for Python mllib PrefixSpan.train(). This stub returns a handle
* to the Java object instead of the content of the Java object. Extra care
* needs to be taken in the Python code to ensure it gets freed on exit; see
* the Py4J documentation.
*/
def trainPrefixSpanModel(
data: JavaRDD[java.util.ArrayList[java.util.ArrayList[Any]]],
minSupport: Double,
maxPatternLength: Int,
localProjDBSize: Int ): PrefixSpanModelWrapper = {
val prefixSpan = new PrefixSpan()
.setMinSupport(minSupport)
.setMaxPatternLength(maxPatternLength)
.setMaxLocalProjDBSize(localProjDBSize)
val trainData = data.rdd.map(_.asScala.toArray.map(_.asScala.toArray))
val model = prefixSpan.run(trainData)
new PrefixSpanModelWrapper(model)
}
/**
* Java stub for Normalizer.transform()
*/
def normalizeVector(p: Double, vector: Vector): Vector = {
new Normalizer(p).transform(vector)
}
/**
* Java stub for Normalizer.transform()
*/
def normalizeVector(p: Double, rdd: JavaRDD[Vector]): JavaRDD[Vector] = {
new Normalizer(p).transform(rdd)
}
/**
* Java stub for StandardScaler.fit(). This stub returns a
* handle to the Java object instead of the content of the Java object.
* Extra care needs to be taken in the Python code to ensure it gets freed on
* exit; see the Py4J documentation.
*/
def fitStandardScaler(
withMean: Boolean,
withStd: Boolean,
data: JavaRDD[Vector]): StandardScalerModel = {
new StandardScaler(withMean, withStd).fit(data.rdd)
}
/**
* Java stub for ChiSqSelector.fit(). This stub returns a
* handle to the Java object instead of the content of the Java object.
* Extra care needs to be taken in the Python code to ensure it gets freed on
* exit; see the Py4J documentation.
*/
def fitChiSqSelector(numTopFeatures: Int, data: JavaRDD[LabeledPoint]): ChiSqSelectorModel = {
new ChiSqSelector(numTopFeatures).fit(data.rdd)
}
/**
* Java stub for PCA.fit(). This stub returns a
* handle to the Java object instead of the content of the Java object.
* Extra care needs to be taken in the Python code to ensure it gets freed on
* exit; see the Py4J documentation.
*/
def fitPCA(k: Int, data: JavaRDD[Vector]): PCAModel = {
new PCA(k).fit(data.rdd)
}
/**
* Java stub for IDF.fit(). This stub returns a
* handle to the Java object instead of the content of the Java object.
* Extra care needs to be taken in the Python code to ensure it gets freed on
* exit; see the Py4J documentation.
*/
def fitIDF(minDocFreq: Int, dataset: JavaRDD[Vector]): IDFModel = {
new IDF(minDocFreq).fit(dataset)
}
/**
* Java stub for Python mllib Word2Vec fit(). This stub returns a
* handle to the Java object instead of the content of the Java object.
* Extra care needs to be taken in the Python code to ensure it gets freed on
* exit; see the Py4J documentation.
* @param dataJRDD input JavaRDD
* @param vectorSize size of vector
* @param learningRate initial learning rate
* @param numPartitions number of partitions
* @param numIterations number of iterations
* @param seed initial seed for random generator
* @return A handle to java Word2VecModelWrapper instance at python side
*/
def trainWord2VecModel(
dataJRDD: JavaRDD[java.util.ArrayList[String]],
vectorSize: Int,
learningRate: Double,
numPartitions: Int,
numIterations: Int,
seed: Long,
minCount: Int): Word2VecModelWrapper = {
val word2vec = new Word2Vec()
.setVectorSize(vectorSize)
.setLearningRate(learningRate)
.setNumPartitions(numPartitions)
.setNumIterations(numIterations)
.setSeed(seed)
.setMinCount(minCount)
try {
val model = word2vec.fit(dataJRDD.rdd.persist(StorageLevel.MEMORY_AND_DISK_SER))
new Word2VecModelWrapper(model)
} finally {
dataJRDD.rdd.unpersist(blocking = false)
}
}
private[python] class Word2VecModelWrapper(model: Word2VecModel) {
def transform(word: String): Vector = {
model.transform(word)
}
/**
* Transforms an RDD of words to its vector representation
* @param rdd an RDD of words
* @return an RDD of vector representations of words
*/
def transform(rdd: JavaRDD[String]): JavaRDD[Vector] = {
rdd.rdd.map(model.transform)
}
def findSynonyms(word: String, num: Int): JList[Object] = {
val vec = transform(word)
findSynonyms(vec, num)
}
def findSynonyms(vector: Vector, num: Int): JList[Object] = {
val result = model.findSynonyms(vector, num)
val similarity = Vectors.dense(result.map(_._2))
val words = result.map(_._1)
List(words, similarity).map(_.asInstanceOf[Object]).asJava
}
def getVectors: JMap[String, JList[Float]] = {
model.getVectors.map({case (k, v) => (k, v.toList.asJava)}).asJava
}
def save(sc: SparkContext, path: String): Unit = model.save(sc, path)
}
/**
* Java stub for Python mllib DecisionTree.train().
* This stub returns a handle to the Java object instead of the content of the Java object.
* Extra care needs to be taken in the Python code to ensure it gets freed on exit;
* see the Py4J documentation.
* @param data Training data
* @param categoricalFeaturesInfo Categorical features info, as Java map
*/
def trainDecisionTreeModel(
data: JavaRDD[LabeledPoint],
algoStr: String,
numClasses: Int,
categoricalFeaturesInfo: JMap[Int, Int],
impurityStr: String,
maxDepth: Int,
maxBins: Int,
minInstancesPerNode: Int,
minInfoGain: Double): DecisionTreeModel = {
val algo = Algo.fromString(algoStr)
val impurity = Impurities.fromString(impurityStr)
val strategy = new Strategy(
algo = algo,
impurity = impurity,
maxDepth = maxDepth,
numClasses = numClasses,
maxBins = maxBins,
categoricalFeaturesInfo = categoricalFeaturesInfo.asScala.toMap,
minInstancesPerNode = minInstancesPerNode,
minInfoGain = minInfoGain)
try {
DecisionTree.train(data.rdd.persist(StorageLevel.MEMORY_AND_DISK), strategy)
} finally {
data.rdd.unpersist(blocking = false)
}
}
/**
* Java stub for Python mllib RandomForest.train().
* This stub returns a handle to the Java object instead of the content of the Java object.
* Extra care needs to be taken in the Python code to ensure it gets freed on exit;
* see the Py4J documentation.
*/
def trainRandomForestModel(
data: JavaRDD[LabeledPoint],
algoStr: String,
numClasses: Int,
categoricalFeaturesInfo: JMap[Int, Int],
numTrees: Int,
featureSubsetStrategy: String,
impurityStr: String,
maxDepth: Int,
maxBins: Int,
seed: Int): RandomForestModel = {
val algo = Algo.fromString(algoStr)
val impurity = Impurities.fromString(impurityStr)
val strategy = new Strategy(
algo = algo,
impurity = impurity,
maxDepth = maxDepth,
numClasses = numClasses,
maxBins = maxBins,
categoricalFeaturesInfo = categoricalFeaturesInfo.asScala.toMap)
val cached = data.rdd.persist(StorageLevel.MEMORY_AND_DISK)
try {
if (algo == Algo.Classification) {
RandomForest.trainClassifier(cached, strategy, numTrees, featureSubsetStrategy, seed)
} else {
RandomForest.trainRegressor(cached, strategy, numTrees, featureSubsetStrategy, seed)
}
} finally {
cached.unpersist(blocking = false)
}
}
/**
* Java stub for Python mllib GradientBoostedTrees.train().
* This stub returns a handle to the Java object instead of the content of the Java object.
* Extra care needs to be taken in the Python code to ensure it gets freed on exit;
* see the Py4J documentation.
*/
def trainGradientBoostedTreesModel(
data: JavaRDD[LabeledPoint],
algoStr: String,
categoricalFeaturesInfo: JMap[Int, Int],
lossStr: String,
numIterations: Int,
learningRate: Double,
maxDepth: Int,
maxBins: Int): GradientBoostedTreesModel = {
val boostingStrategy = BoostingStrategy.defaultParams(algoStr)
boostingStrategy.setLoss(Losses.fromString(lossStr))
boostingStrategy.setNumIterations(numIterations)
boostingStrategy.setLearningRate(learningRate)
boostingStrategy.treeStrategy.setMaxDepth(maxDepth)
boostingStrategy.treeStrategy.setMaxBins(maxBins)
boostingStrategy.treeStrategy.categoricalFeaturesInfo = categoricalFeaturesInfo.asScala.toMap
val cached = data.rdd.persist(StorageLevel.MEMORY_AND_DISK)
try {
GradientBoostedTrees.train(cached, boostingStrategy)
} finally {
cached.unpersist(blocking = false)
}
}
def elementwiseProductVector(scalingVector: Vector, vector: Vector): Vector = {
new ElementwiseProduct(scalingVector).transform(vector)
}
def elementwiseProductVector(scalingVector: Vector, vector: JavaRDD[Vector]): JavaRDD[Vector] = {
new ElementwiseProduct(scalingVector).transform(vector)
}
/**
* Java stub for mllib Statistics.colStats(X: RDD[Vector]).
* TODO figure out return type.
*/
def colStats(rdd: JavaRDD[Vector]): MultivariateStatisticalSummary = {
Statistics.colStats(rdd.rdd)
}
/**
* Java stub for mllib Statistics.corr(X: RDD[Vector], method: String).
* Returns the correlation matrix serialized into a byte array understood by deserializers in
* pyspark.
*/
def corr(x: JavaRDD[Vector], method: String): Matrix = {
Statistics.corr(x.rdd, getCorrNameOrDefault(method))
}
/**
* Java stub for mllib Statistics.corr(x: RDD[Double], y: RDD[Double], method: String).
*/
def corr(x: JavaRDD[Double], y: JavaRDD[Double], method: String): Double = {
Statistics.corr(x.rdd, y.rdd, getCorrNameOrDefault(method))
}
/**
* Java stub for mllib Statistics.chiSqTest()
*/
def chiSqTest(observed: Vector, expected: Vector): ChiSqTestResult = {
if (expected == null) {
Statistics.chiSqTest(observed)
} else {
Statistics.chiSqTest(observed, expected)
}
}
/**
* Java stub for mllib Statistics.chiSqTest(observed: Matrix)
*/
def chiSqTest(observed: Matrix): ChiSqTestResult = {
Statistics.chiSqTest(observed)
}
/**
* Java stub for mllib Statistics.chiSqTest(RDD[LabelPoint])
*/
def chiSqTest(data: JavaRDD[LabeledPoint]): Array[ChiSqTestResult] = {
Statistics.chiSqTest(data.rdd)
}
// used by the corr methods to retrieve the name of the correlation method passed in via pyspark
private def getCorrNameOrDefault(method: String) = {
if (method == null) CorrelationNames.defaultCorrName else method
}
// Used by the *RDD methods to get default seed if not passed in from pyspark
private def getSeedOrDefault(seed: java.lang.Long): Long = {
if (seed == null) Utils.random.nextLong else seed
}
// Used by *RDD methods to get default numPartitions if not passed in from pyspark
private def getNumPartitionsOrDefault(numPartitions: java.lang.Integer,
jsc: JavaSparkContext): Int = {
if (numPartitions == null) {
jsc.sc.defaultParallelism
} else {
numPartitions
}
}
// Note: for the following methods, numPartitions and seed are boxed to allow nulls to be passed
// in for either argument from pyspark
/**
* Java stub for Python mllib RandomRDDGenerators.uniformRDD()
*/
def uniformRDD(jsc: JavaSparkContext,
size: Long,
numPartitions: java.lang.Integer,
seed: java.lang.Long): JavaRDD[Double] = {
val parts = getNumPartitionsOrDefault(numPartitions, jsc)
val s = getSeedOrDefault(seed)
RG.uniformRDD(jsc.sc, size, parts, s)
}
/**
* Java stub for Python mllib RandomRDDGenerators.normalRDD()
*/
def normalRDD(jsc: JavaSparkContext,
size: Long,
numPartitions: java.lang.Integer,
seed: java.lang.Long): JavaRDD[Double] = {
val parts = getNumPartitionsOrDefault(numPartitions, jsc)
val s = getSeedOrDefault(seed)
RG.normalRDD(jsc.sc, size, parts, s)
}
/**
* Java stub for Python mllib RandomRDDGenerators.logNormalRDD()
*/
def logNormalRDD(jsc: JavaSparkContext,
mean: Double,
std: Double,
size: Long,
numPartitions: java.lang.Integer,
seed: java.lang.Long): JavaRDD[Double] = {
val parts = getNumPartitionsOrDefault(numPartitions, jsc)
val s = getSeedOrDefault(seed)
RG.logNormalRDD(jsc.sc, mean, std, size, parts, s)
}
/**
* Java stub for Python mllib RandomRDDGenerators.poissonRDD()
*/
def poissonRDD(jsc: JavaSparkContext,
mean: Double,
size: Long,
numPartitions: java.lang.Integer,
seed: java.lang.Long): JavaRDD[Double] = {
val parts = getNumPartitionsOrDefault(numPartitions, jsc)
val s = getSeedOrDefault(seed)
RG.poissonRDD(jsc.sc, mean, size, parts, s)
}
/**
* Java stub for Python mllib RandomRDDGenerators.exponentialRDD()
*/
def exponentialRDD(jsc: JavaSparkContext,
mean: Double,
size: Long,
numPartitions: java.lang.Integer,
seed: java.lang.Long): JavaRDD[Double] = {
val parts = getNumPartitionsOrDefault(numPartitions, jsc)
val s = getSeedOrDefault(seed)
RG.exponentialRDD(jsc.sc, mean, size, parts, s)
}
/**
* Java stub for Python mllib RandomRDDGenerators.gammaRDD()
*/
def gammaRDD(jsc: JavaSparkContext,
shape: Double,
scale: Double,
size: Long,
numPartitions: java.lang.Integer,
seed: java.lang.Long): JavaRDD[Double] = {
val parts = getNumPartitionsOrDefault(numPartitions, jsc)
val s = getSeedOrDefault(seed)
RG.gammaRDD(jsc.sc, shape, scale, size, parts, s)
}
/**
* Java stub for Python mllib RandomRDDGenerators.uniformVectorRDD()
*/
def uniformVectorRDD(jsc: JavaSparkContext,
numRows: Long,
numCols: Int,
numPartitions: java.lang.Integer,
seed: java.lang.Long): JavaRDD[Vector] = {
val parts = getNumPartitionsOrDefault(numPartitions, jsc)
val s = getSeedOrDefault(seed)
RG.uniformVectorRDD(jsc.sc, numRows, numCols, parts, s)
}
/**
* Java stub for Python mllib RandomRDDGenerators.normalVectorRDD()
*/
def normalVectorRDD(jsc: JavaSparkContext,
numRows: Long,
numCols: Int,
numPartitions: java.lang.Integer,
seed: java.lang.Long): JavaRDD[Vector] = {
val parts = getNumPartitionsOrDefault(numPartitions, jsc)
val s = getSeedOrDefault(seed)
RG.normalVectorRDD(jsc.sc, numRows, numCols, parts, s)
}
/**
* Java stub for Python mllib RandomRDDGenerators.logNormalVectorRDD()
*/
def logNormalVectorRDD(jsc: JavaSparkContext,
mean: Double,
std: Double,
numRows: Long,
numCols: Int,
numPartitions: java.lang.Integer,
seed: java.lang.Long): JavaRDD[Vector] = {
val parts = getNumPartitionsOrDefault(numPartitions, jsc)
val s = getSeedOrDefault(seed)
RG.logNormalVectorRDD(jsc.sc, mean, std, numRows, numCols, parts, s)
}
/**
* Java stub for Python mllib RandomRDDGenerators.poissonVectorRDD()
*/
def poissonVectorRDD(jsc: JavaSparkContext,
mean: Double,
numRows: Long,
numCols: Int,
numPartitions: java.lang.Integer,
seed: java.lang.Long): JavaRDD[Vector] = {
val parts = getNumPartitionsOrDefault(numPartitions, jsc)
val s = getSeedOrDefault(seed)
RG.poissonVectorRDD(jsc.sc, mean, numRows, numCols, parts, s)
}
/**
* Java stub for Python mllib RandomRDDGenerators.exponentialVectorRDD()
*/
def exponentialVectorRDD(jsc: JavaSparkContext,
mean: Double,
numRows: Long,
numCols: Int,
numPartitions: java.lang.Integer,
seed: java.lang.Long): JavaRDD[Vector] = {
val parts = getNumPartitionsOrDefault(numPartitions, jsc)
val s = getSeedOrDefault(seed)
RG.exponentialVectorRDD(jsc.sc, mean, numRows, numCols, parts, s)
}
/**
* Java stub for Python mllib RandomRDDGenerators.gammaVectorRDD()
*/
def gammaVectorRDD(jsc: JavaSparkContext,
shape: Double,
scale: Double,
numRows: Long,
numCols: Int,
numPartitions: java.lang.Integer,
seed: java.lang.Long): JavaRDD[Vector] = {
val parts = getNumPartitionsOrDefault(numPartitions, jsc)
val s = getSeedOrDefault(seed)
RG.gammaVectorRDD(jsc.sc, shape, scale, numRows, numCols, parts, s)
}
/**
* Java stub for the constructor of Python mllib RankingMetrics
*/
def newRankingMetrics(predictionAndLabels: DataFrame): RankingMetrics[Any] = {
new RankingMetrics(predictionAndLabels.map(
r => (r.getSeq(0).toArray[Any], r.getSeq(1).toArray[Any])))
}
/**
* Java stub for the estimate method of KernelDensity
*/
def estimateKernelDensity(
sample: JavaRDD[Double],
bandwidth: Double, points: java.util.ArrayList[Double]): Array[Double] = {
new KernelDensity().setSample(sample).setBandwidth(bandwidth).estimate(
points.asScala.toArray)
}
/**
* Java stub for the update method of StreamingKMeansModel.
*/
def updateStreamingKMeansModel(
clusterCenters: JList[Vector],
clusterWeights: JList[Double],
data: JavaRDD[Vector],
decayFactor: Double,
timeUnit: String): JList[Object] = {
val model = new StreamingKMeansModel(
clusterCenters.asScala.toArray, clusterWeights.asScala.toArray)
.update(data, decayFactor, timeUnit)
List[AnyRef](model.clusterCenters, Vectors.dense(model.clusterWeights)).asJava
}
/**
* Wrapper around the generateLinearInput method of LinearDataGenerator.
*/
def generateLinearInputWrapper(
intercept: Double,
weights: JList[Double],
xMean: JList[Double],
xVariance: JList[Double],
nPoints: Int,
seed: Int,
eps: Double): Array[LabeledPoint] = {
LinearDataGenerator.generateLinearInput(
intercept, weights.asScala.toArray, xMean.asScala.toArray,
xVariance.asScala.toArray, nPoints, seed, eps).toArray
}
/**
* Wrapper around the generateLinearRDD method of LinearDataGenerator.
*/
def generateLinearRDDWrapper(
sc: JavaSparkContext,
nexamples: Int,
nfeatures: Int,
eps: Double,
nparts: Int,
intercept: Double): JavaRDD[LabeledPoint] = {
LinearDataGenerator.generateLinearRDD(
sc, nexamples, nfeatures, eps, nparts, intercept)
}
/**
* Java stub for Statistics.kolmogorovSmirnovTest()
*/
def kolmogorovSmirnovTest(
data: JavaRDD[Double],
distName: String,
params: JList[Double]): KolmogorovSmirnovTestResult = {
val paramsSeq = params.asScala.toSeq
Statistics.kolmogorovSmirnovTest(data, distName, paramsSeq: _*)
}
/**
* Wrapper around RowMatrix constructor.
*/
def createRowMatrix(rows: JavaRDD[Vector], numRows: Long, numCols: Int): RowMatrix = {
new RowMatrix(rows.rdd, numRows, numCols)
}
/**
* Wrapper around IndexedRowMatrix constructor.
*/
def createIndexedRowMatrix(rows: DataFrame, numRows: Long, numCols: Int): IndexedRowMatrix = {
// We use DataFrames for serialization of IndexedRows from Python,
// so map each Row in the DataFrame back to an IndexedRow.
val indexedRows = rows.map {
case Row(index: Long, vector: Vector) => IndexedRow(index, vector)
}
new IndexedRowMatrix(indexedRows, numRows, numCols)
}
/**
* Wrapper around CoordinateMatrix constructor.
*/
def createCoordinateMatrix(rows: DataFrame, numRows: Long, numCols: Long): CoordinateMatrix = {
// We use DataFrames for serialization of MatrixEntry entries from
// Python, so map each Row in the DataFrame back to a MatrixEntry.
val entries = rows.map {
case Row(i: Long, j: Long, value: Double) => MatrixEntry(i, j, value)
}
new CoordinateMatrix(entries, numRows, numCols)
}
/**
* Wrapper around BlockMatrix constructor.
*/
def createBlockMatrix(blocks: DataFrame, rowsPerBlock: Int, colsPerBlock: Int,
numRows: Long, numCols: Long): BlockMatrix = {
// We use DataFrames for serialization of sub-matrix blocks from
// Python, so map each Row in the DataFrame back to a
// ((blockRowIndex, blockColIndex), sub-matrix) tuple.
val blockTuples = blocks.map {
case Row(Row(blockRowIndex: Long, blockColIndex: Long), subMatrix: Matrix) =>
((blockRowIndex.toInt, blockColIndex.toInt), subMatrix)
}
new BlockMatrix(blockTuples, rowsPerBlock, colsPerBlock, numRows, numCols)
}
/**
* Return the rows of an IndexedRowMatrix.
*/
def getIndexedRows(indexedRowMatrix: IndexedRowMatrix): DataFrame = {
// We use DataFrames for serialization of IndexedRows to Python,
// so return a DataFrame.
val sqlContext = SQLContext.getOrCreate(indexedRowMatrix.rows.sparkContext)
sqlContext.createDataFrame(indexedRowMatrix.rows)
}
/**
* Return the entries of a CoordinateMatrix.
*/
def getMatrixEntries(coordinateMatrix: CoordinateMatrix): DataFrame = {
// We use DataFrames for serialization of MatrixEntry entries to
// Python, so return a DataFrame.
val sqlContext = SQLContext.getOrCreate(coordinateMatrix.entries.sparkContext)
sqlContext.createDataFrame(coordinateMatrix.entries)
}
/**
* Return the sub-matrix blocks of a BlockMatrix.
*/
def getMatrixBlocks(blockMatrix: BlockMatrix): DataFrame = {
// We use DataFrames for serialization of sub-matrix blocks to
// Python, so return a DataFrame.
val sqlContext = SQLContext.getOrCreate(blockMatrix.blocks.sparkContext)
sqlContext.createDataFrame(blockMatrix.blocks)
}
}
/**
* SerDe utility functions for PythonMLLibAPI.
*/
private[spark] object SerDe extends Serializable {
val PYSPARK_PACKAGE = "pyspark.mllib"
val LATIN1 = "ISO-8859-1"
/**
* Base class used for pickle
*/
private[python] abstract class BasePickler[T: ClassTag]
extends IObjectPickler with IObjectConstructor {
private val cls = implicitly[ClassTag[T]].runtimeClass
private val module = PYSPARK_PACKAGE + "." + cls.getName.split('.')(4)
private val name = cls.getSimpleName
// register this to Pickler and Unpickler
def register(): Unit = {
Pickler.registerCustomPickler(this.getClass, this)
Pickler.registerCustomPickler(cls, this)
Unpickler.registerConstructor(module, name, this)
}
def pickle(obj: Object, out: OutputStream, pickler: Pickler): Unit = {
if (obj == this) {
out.write(Opcodes.GLOBAL)
out.write((module + "\n" + name + "\n").getBytes)
} else {
pickler.save(this) // it will be memorized by Pickler
saveState(obj, out, pickler)
out.write(Opcodes.REDUCE)
}
}
private[python] def saveObjects(out: OutputStream, pickler: Pickler, objects: Any*) = {
if (objects.length == 0 || objects.length > 3) {
out.write(Opcodes.MARK)
}
objects.foreach(pickler.save)
val code = objects.length match {
case 1 => Opcodes.TUPLE1
case 2 => Opcodes.TUPLE2
case 3 => Opcodes.TUPLE3
case _ => Opcodes.TUPLE
}
out.write(code)
}
protected def getBytes(obj: Object): Array[Byte] = {
if (obj.getClass.isArray) {
obj.asInstanceOf[Array[Byte]]
} else {
obj.asInstanceOf[String].getBytes(LATIN1)
}
}
private[python] def saveState(obj: Object, out: OutputStream, pickler: Pickler)
}
// Pickler for DenseVector
private[python] class DenseVectorPickler extends BasePickler[DenseVector] {
def saveState(obj: Object, out: OutputStream, pickler: Pickler): Unit = {
val vector: DenseVector = obj.asInstanceOf[DenseVector]
val bytes = new Array[Byte](8 * vector.size)
val bb = ByteBuffer.wrap(bytes)
bb.order(ByteOrder.nativeOrder())
val db = bb.asDoubleBuffer()
db.put(vector.values)
out.write(Opcodes.BINSTRING)
out.write(PickleUtils.integer_to_bytes(bytes.length))
out.write(bytes)
out.write(Opcodes.TUPLE1)
}
def construct(args: Array[Object]): Object = {
require(args.length == 1)
if (args.length != 1) {
throw new PickleException("should be 1")
}
val bytes = getBytes(args(0))
val bb = ByteBuffer.wrap(bytes, 0, bytes.length)
bb.order(ByteOrder.nativeOrder())
val db = bb.asDoubleBuffer()
val ans = new Array[Double](bytes.length / 8)
db.get(ans)
Vectors.dense(ans)
}
}
// Pickler for DenseMatrix
private[python] class DenseMatrixPickler extends BasePickler[DenseMatrix] {
def saveState(obj: Object, out: OutputStream, pickler: Pickler): Unit = {
val m: DenseMatrix = obj.asInstanceOf[DenseMatrix]
val bytes = new Array[Byte](8 * m.values.size)
val order = ByteOrder.nativeOrder()
val isTransposed = if (m.isTransposed) 1 else 0
ByteBuffer.wrap(bytes).order(order).asDoubleBuffer().put(m.values)
out.write(Opcodes.MARK)
out.write(Opcodes.BININT)
out.write(PickleUtils.integer_to_bytes(m.numRows))
out.write(Opcodes.BININT)
out.write(PickleUtils.integer_to_bytes(m.numCols))
out.write(Opcodes.BINSTRING)
out.write(PickleUtils.integer_to_bytes(bytes.length))
out.write(bytes)
out.write(Opcodes.BININT)
out.write(PickleUtils.integer_to_bytes(isTransposed))
out.write(Opcodes.TUPLE)
}
def construct(args: Array[Object]): Object = {
if (args.length != 4) {
throw new PickleException("should be 4")
}
val bytes = getBytes(args(2))
val n = bytes.length / 8
val values = new Array[Double](n)
val order = ByteOrder.nativeOrder()
ByteBuffer.wrap(bytes).order(order).asDoubleBuffer().get(values)
val isTransposed = args(3).asInstanceOf[Int] == 1
new DenseMatrix(args(0).asInstanceOf[Int], args(1).asInstanceOf[Int], values, isTransposed)
}
}
// Pickler for SparseMatrix
private[python] class SparseMatrixPickler extends BasePickler[SparseMatrix] {
def saveState(obj: Object, out: OutputStream, pickler: Pickler): Unit = {
val s = obj.asInstanceOf[SparseMatrix]
val order = ByteOrder.nativeOrder()
val colPtrsBytes = new Array[Byte](4 * s.colPtrs.length)
val indicesBytes = new Array[Byte](4 * s.rowIndices.length)
val valuesBytes = new Array[Byte](8 * s.values.length)
val isTransposed = if (s.isTransposed) 1 else 0
ByteBuffer.wrap(colPtrsBytes).order(order).asIntBuffer().put(s.colPtrs)
ByteBuffer.wrap(indicesBytes).order(order).asIntBuffer().put(s.rowIndices)
ByteBuffer.wrap(valuesBytes).order(order).asDoubleBuffer().put(s.values)
out.write(Opcodes.MARK)
out.write(Opcodes.BININT)
out.write(PickleUtils.integer_to_bytes(s.numRows))
out.write(Opcodes.BININT)
out.write(PickleUtils.integer_to_bytes(s.numCols))
out.write(Opcodes.BINSTRING)
out.write(PickleUtils.integer_to_bytes(colPtrsBytes.length))
out.write(colPtrsBytes)
out.write(Opcodes.BINSTRING)
out.write(PickleUtils.integer_to_bytes(indicesBytes.length))
out.write(indicesBytes)
out.write(Opcodes.BINSTRING)
out.write(PickleUtils.integer_to_bytes(valuesBytes.length))
out.write(valuesBytes)
out.write(Opcodes.BININT)
out.write(PickleUtils.integer_to_bytes(isTransposed))
out.write(Opcodes.TUPLE)
}
def construct(args: Array[Object]): Object = {
if (args.length != 6) {
throw new PickleException("should be 6")
}
val order = ByteOrder.nativeOrder()
val colPtrsBytes = getBytes(args(2))
val indicesBytes = getBytes(args(3))
val valuesBytes = getBytes(args(4))
val colPtrs = new Array[Int](colPtrsBytes.length / 4)
val rowIndices = new Array[Int](indicesBytes.length / 4)
val values = new Array[Double](valuesBytes.length / 8)
ByteBuffer.wrap(colPtrsBytes).order(order).asIntBuffer().get(colPtrs)
ByteBuffer.wrap(indicesBytes).order(order).asIntBuffer().get(rowIndices)
ByteBuffer.wrap(valuesBytes).order(order).asDoubleBuffer().get(values)
val isTransposed = args(5).asInstanceOf[Int] == 1
new SparseMatrix(
args(0).asInstanceOf[Int], args(1).asInstanceOf[Int], colPtrs, rowIndices, values,
isTransposed)
}
}
// Pickler for SparseVector
private[python] class SparseVectorPickler extends BasePickler[SparseVector] {
def saveState(obj: Object, out: OutputStream, pickler: Pickler): Unit = {
val v: SparseVector = obj.asInstanceOf[SparseVector]
val n = v.indices.size
val indiceBytes = new Array[Byte](4 * n)
val order = ByteOrder.nativeOrder()
ByteBuffer.wrap(indiceBytes).order(order).asIntBuffer().put(v.indices)
val valueBytes = new Array[Byte](8 * n)
ByteBuffer.wrap(valueBytes).order(order).asDoubleBuffer().put(v.values)
out.write(Opcodes.BININT)
out.write(PickleUtils.integer_to_bytes(v.size))
out.write(Opcodes.BINSTRING)
out.write(PickleUtils.integer_to_bytes(indiceBytes.length))
out.write(indiceBytes)
out.write(Opcodes.BINSTRING)
out.write(PickleUtils.integer_to_bytes(valueBytes.length))
out.write(valueBytes)
out.write(Opcodes.TUPLE3)
}
def construct(args: Array[Object]): Object = {
if (args.length != 3) {
throw new PickleException("should be 3")
}
val size = args(0).asInstanceOf[Int]
val indiceBytes = getBytes(args(1))
val valueBytes = getBytes(args(2))
val n = indiceBytes.length / 4
val indices = new Array[Int](n)
val values = new Array[Double](n)
if (n > 0) {
val order = ByteOrder.nativeOrder()
ByteBuffer.wrap(indiceBytes).order(order).asIntBuffer().get(indices)
ByteBuffer.wrap(valueBytes).order(order).asDoubleBuffer().get(values)
}
new SparseVector(size, indices, values)
}
}
// Pickler for LabeledPoint
private[python] class LabeledPointPickler extends BasePickler[LabeledPoint] {
def saveState(obj: Object, out: OutputStream, pickler: Pickler): Unit = {
val point: LabeledPoint = obj.asInstanceOf[LabeledPoint]
saveObjects(out, pickler, point.label, point.features)
}
def construct(args: Array[Object]): Object = {
if (args.length != 2) {
throw new PickleException("should be 2")
}
new LabeledPoint(args(0).asInstanceOf[Double], args(1).asInstanceOf[Vector])
}
}
// Pickler for Rating
private[python] class RatingPickler extends BasePickler[Rating] {
def saveState(obj: Object, out: OutputStream, pickler: Pickler): Unit = {
val rating: Rating = obj.asInstanceOf[Rating]
saveObjects(out, pickler, rating.user, rating.product, rating.rating)
}
def construct(args: Array[Object]): Object = {
if (args.length != 3) {
throw new PickleException("should be 3")
}
new Rating(args(0).asInstanceOf[Int], args(1).asInstanceOf[Int],
args(2).asInstanceOf[Double])
}
}
var initialized = false
// This should be called before trying to serialize any above classes
// In cluster mode, this should be put in the closure
def initialize(): Unit = {
SerDeUtil.initialize()
synchronized {
if (!initialized) {
new DenseVectorPickler().register()
new DenseMatrixPickler().register()
new SparseMatrixPickler().register()
new SparseVectorPickler().register()
new LabeledPointPickler().register()
new RatingPickler().register()
initialized = true
}
}
}
// will not called in Executor automatically
initialize()
def dumps(obj: AnyRef): Array[Byte] = {
new Pickler().dumps(obj)
}
def loads(bytes: Array[Byte]): AnyRef = {
new Unpickler().loads(bytes)
}
/* convert object into Tuple */
def asTupleRDD(rdd: RDD[Array[Any]]): RDD[(Int, Int)] = {
rdd.map(x => (x(0).asInstanceOf[Int], x(1).asInstanceOf[Int]))
}
/* convert RDD[Tuple2[,]] to RDD[Array[Any]] */
def fromTuple2RDD(rdd: RDD[(Any, Any)]): RDD[Array[Any]] = {
rdd.map(x => Array(x._1, x._2))
}
/**
* Convert an RDD of Java objects to an RDD of serialized Python objects, that is usable by
* PySpark.
*/
def javaToPython(jRDD: JavaRDD[Any]): JavaRDD[Array[Byte]] = {
jRDD.rdd.mapPartitions { iter =>
initialize() // let it called in executor
new SerDeUtil.AutoBatchedPickler(iter)
}
}
/**
* Convert an RDD of serialized Python objects to RDD of objects, that is usable by PySpark.
*/
def pythonToJava(pyRDD: JavaRDD[Array[Byte]], batched: Boolean): JavaRDD[Any] = {
pyRDD.rdd.mapPartitions { iter =>
initialize() // let it called in executor
val unpickle = new Unpickler
iter.flatMap { row =>
val obj = unpickle.loads(row)
if (batched) {
obj match {
case list: JArrayList[_] => list.asScala
case arr: Array[_] => arr
}
} else {
Seq(obj)
}
}
}.toJavaRDD()
}
}
|