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
|
#
# 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.
#
import itertools
import numpy as np
from pyspark import SparkContext
from pyspark import since, keyword_only
from pyspark.ml import Estimator, Model
from pyspark.ml.param import Params, Param, TypeConverters
from pyspark.ml.param.shared import HasSeed
from pyspark.ml.wrapper import JavaParams
from pyspark.sql.functions import rand
from pyspark.mllib.common import inherit_doc, _py2java
__all__ = ['ParamGridBuilder', 'CrossValidator', 'CrossValidatorModel', 'TrainValidationSplit',
'TrainValidationSplitModel']
class ParamGridBuilder(object):
r"""
.. note:: Experimental
Builder for a param grid used in grid search-based model selection.
>>> from pyspark.ml.classification import LogisticRegression
>>> lr = LogisticRegression()
>>> output = ParamGridBuilder() \
... .baseOn({lr.labelCol: 'l'}) \
... .baseOn([lr.predictionCol, 'p']) \
... .addGrid(lr.regParam, [1.0, 2.0]) \
... .addGrid(lr.maxIter, [1, 5]) \
... .build()
>>> expected = [
... {lr.regParam: 1.0, lr.maxIter: 1, lr.labelCol: 'l', lr.predictionCol: 'p'},
... {lr.regParam: 2.0, lr.maxIter: 1, lr.labelCol: 'l', lr.predictionCol: 'p'},
... {lr.regParam: 1.0, lr.maxIter: 5, lr.labelCol: 'l', lr.predictionCol: 'p'},
... {lr.regParam: 2.0, lr.maxIter: 5, lr.labelCol: 'l', lr.predictionCol: 'p'}]
>>> len(output) == len(expected)
True
>>> all([m in expected for m in output])
True
.. versionadded:: 1.4.0
"""
def __init__(self):
self._param_grid = {}
@since("1.4.0")
def addGrid(self, param, values):
"""
Sets the given parameters in this grid to fixed values.
"""
self._param_grid[param] = values
return self
@since("1.4.0")
def baseOn(self, *args):
"""
Sets the given parameters in this grid to fixed values.
Accepts either a parameter dictionary or a list of (parameter, value) pairs.
"""
if isinstance(args[0], dict):
self.baseOn(*args[0].items())
else:
for (param, value) in args:
self.addGrid(param, [value])
return self
@since("1.4.0")
def build(self):
"""
Builds and returns all combinations of parameters specified
by the param grid.
"""
keys = self._param_grid.keys()
grid_values = self._param_grid.values()
return [dict(zip(keys, prod)) for prod in itertools.product(*grid_values)]
class ValidatorParams(HasSeed):
"""
Common params for TrainValidationSplit and CrossValidator.
"""
estimator = Param(Params._dummy(), "estimator", "estimator to be cross-validated")
estimatorParamMaps = Param(Params._dummy(), "estimatorParamMaps", "estimator param maps")
evaluator = Param(
Params._dummy(), "evaluator",
"evaluator used to select hyper-parameters that maximize the validator metric")
def setEstimator(self, value):
"""
Sets the value of :py:attr:`estimator`.
"""
return self._set(estimator=value)
def getEstimator(self):
"""
Gets the value of estimator or its default value.
"""
return self.getOrDefault(self.estimator)
def setEstimatorParamMaps(self, value):
"""
Sets the value of :py:attr:`estimatorParamMaps`.
"""
return self._set(estimatorParamMaps=value)
def getEstimatorParamMaps(self):
"""
Gets the value of estimatorParamMaps or its default value.
"""
return self.getOrDefault(self.estimatorParamMaps)
def setEvaluator(self, value):
"""
Sets the value of :py:attr:`evaluator`.
"""
return self._set(evaluator=value)
def getEvaluator(self):
"""
Gets the value of evaluator or its default value.
"""
return self.getOrDefault(self.evaluator)
class CrossValidator(Estimator, ValidatorParams):
"""
.. note:: Experimental
K-fold cross validation.
>>> from pyspark.ml.classification import LogisticRegression
>>> from pyspark.ml.evaluation import BinaryClassificationEvaluator
>>> from pyspark.ml.linalg import Vectors
>>> dataset = sqlContext.createDataFrame(
... [(Vectors.dense([0.0]), 0.0),
... (Vectors.dense([0.4]), 1.0),
... (Vectors.dense([0.5]), 0.0),
... (Vectors.dense([0.6]), 1.0),
... (Vectors.dense([1.0]), 1.0)] * 10,
... ["features", "label"])
>>> lr = LogisticRegression()
>>> grid = ParamGridBuilder().addGrid(lr.maxIter, [0, 1]).build()
>>> evaluator = BinaryClassificationEvaluator()
>>> cv = CrossValidator(estimator=lr, estimatorParamMaps=grid, evaluator=evaluator)
>>> cvModel = cv.fit(dataset)
>>> evaluator.evaluate(cvModel.transform(dataset))
0.8333...
.. versionadded:: 1.4.0
"""
numFolds = Param(Params._dummy(), "numFolds", "number of folds for cross validation",
typeConverter=TypeConverters.toInt)
@keyword_only
def __init__(self, estimator=None, estimatorParamMaps=None, evaluator=None, numFolds=3,
seed=None):
"""
__init__(self, estimator=None, estimatorParamMaps=None, evaluator=None, numFolds=3,\
seed=None)
"""
super(CrossValidator, self).__init__()
self._setDefault(numFolds=3)
kwargs = self.__init__._input_kwargs
self._set(**kwargs)
@keyword_only
@since("1.4.0")
def setParams(self, estimator=None, estimatorParamMaps=None, evaluator=None, numFolds=3,
seed=None):
"""
setParams(self, estimator=None, estimatorParamMaps=None, evaluator=None, numFolds=3,\
seed=None):
Sets params for cross validator.
"""
kwargs = self.setParams._input_kwargs
return self._set(**kwargs)
@since("1.4.0")
def setNumFolds(self, value):
"""
Sets the value of :py:attr:`numFolds`.
"""
return self._set(numFolds=value)
@since("1.4.0")
def getNumFolds(self):
"""
Gets the value of numFolds or its default value.
"""
return self.getOrDefault(self.numFolds)
def _fit(self, dataset):
est = self.getOrDefault(self.estimator)
epm = self.getOrDefault(self.estimatorParamMaps)
numModels = len(epm)
eva = self.getOrDefault(self.evaluator)
nFolds = self.getOrDefault(self.numFolds)
seed = self.getOrDefault(self.seed)
h = 1.0 / nFolds
randCol = self.uid + "_rand"
df = dataset.select("*", rand(seed).alias(randCol))
metrics = [0.0] * numModels
for i in range(nFolds):
validateLB = i * h
validateUB = (i + 1) * h
condition = (df[randCol] >= validateLB) & (df[randCol] < validateUB)
validation = df.filter(condition)
train = df.filter(~condition)
for j in range(numModels):
model = est.fit(train, epm[j])
# TODO: duplicate evaluator to take extra params from input
metric = eva.evaluate(model.transform(validation, epm[j]))
metrics[j] += metric
if eva.isLargerBetter():
bestIndex = np.argmax(metrics)
else:
bestIndex = np.argmin(metrics)
bestModel = est.fit(dataset, epm[bestIndex])
return self._copyValues(CrossValidatorModel(bestModel, metrics))
@since("1.4.0")
def copy(self, extra=None):
"""
Creates a copy of this instance with a randomly generated uid
and some extra params. This copies creates a deep copy of
the embedded paramMap, and copies the embedded and extra parameters over.
:param extra: Extra parameters to copy to the new instance
:return: Copy of this instance
"""
if extra is None:
extra = dict()
newCV = Params.copy(self, extra)
if self.isSet(self.estimator):
newCV.setEstimator(self.getEstimator().copy(extra))
# estimatorParamMaps remain the same
if self.isSet(self.evaluator):
newCV.setEvaluator(self.getEvaluator().copy(extra))
return newCV
class CrossValidatorModel(Model, ValidatorParams):
"""
.. note:: Experimental
Model from k-fold cross validation.
.. versionadded:: 1.4.0
"""
def __init__(self, bestModel, avgMetrics=[]):
super(CrossValidatorModel, self).__init__()
#: best model from cross validation
self.bestModel = bestModel
#: Average cross-validation metrics for each paramMap in
#: CrossValidator.estimatorParamMaps, in the corresponding order.
self.avgMetrics = avgMetrics
def _transform(self, dataset):
return self.bestModel.transform(dataset)
@since("1.4.0")
def copy(self, extra=None):
"""
Creates a copy of this instance with a randomly generated uid
and some extra params. This copies the underlying bestModel,
creates a deep copy of the embedded paramMap, and
copies the embedded and extra parameters over.
:param extra: Extra parameters to copy to the new instance
:return: Copy of this instance
"""
if extra is None:
extra = dict()
bestModel = self.bestModel.copy(extra)
avgMetrics = self.avgMetrics
return CrossValidatorModel(bestModel, avgMetrics)
class TrainValidationSplit(Estimator, ValidatorParams):
"""
.. note:: Experimental
Validation for hyper-parameter tuning. Randomly splits the input dataset into train and
validation sets, and uses evaluation metric on the validation set to select the best model.
Similar to :class:`CrossValidator`, but only splits the set once.
>>> from pyspark.ml.classification import LogisticRegression
>>> from pyspark.ml.evaluation import BinaryClassificationEvaluator
>>> from pyspark.ml.linalg import Vectors
>>> dataset = sqlContext.createDataFrame(
... [(Vectors.dense([0.0]), 0.0),
... (Vectors.dense([0.4]), 1.0),
... (Vectors.dense([0.5]), 0.0),
... (Vectors.dense([0.6]), 1.0),
... (Vectors.dense([1.0]), 1.0)] * 10,
... ["features", "label"])
>>> lr = LogisticRegression()
>>> grid = ParamGridBuilder().addGrid(lr.maxIter, [0, 1]).build()
>>> evaluator = BinaryClassificationEvaluator()
>>> tvs = TrainValidationSplit(estimator=lr, estimatorParamMaps=grid, evaluator=evaluator)
>>> tvsModel = tvs.fit(dataset)
>>> evaluator.evaluate(tvsModel.transform(dataset))
0.8333...
.. versionadded:: 2.0.0
"""
trainRatio = Param(Params._dummy(), "trainRatio", "Param for ratio between train and\
validation data. Must be between 0 and 1.", typeConverter=TypeConverters.toFloat)
@keyword_only
def __init__(self, estimator=None, estimatorParamMaps=None, evaluator=None, trainRatio=0.75,
seed=None):
"""
__init__(self, estimator=None, estimatorParamMaps=None, evaluator=None, trainRatio=0.75,\
seed=None)
"""
super(TrainValidationSplit, self).__init__()
self._setDefault(trainRatio=0.75)
kwargs = self.__init__._input_kwargs
self._set(**kwargs)
@since("2.0.0")
@keyword_only
def setParams(self, estimator=None, estimatorParamMaps=None, evaluator=None, trainRatio=0.75,
seed=None):
"""
setParams(self, estimator=None, estimatorParamMaps=None, evaluator=None, trainRatio=0.75,\
seed=None):
Sets params for the train validation split.
"""
kwargs = self.setParams._input_kwargs
return self._set(**kwargs)
@since("2.0.0")
def setTrainRatio(self, value):
"""
Sets the value of :py:attr:`trainRatio`.
"""
return self._set(trainRatio=value)
@since("2.0.0")
def getTrainRatio(self):
"""
Gets the value of trainRatio or its default value.
"""
return self.getOrDefault(self.trainRatio)
def _fit(self, dataset):
est = self.getOrDefault(self.estimator)
epm = self.getOrDefault(self.estimatorParamMaps)
numModels = len(epm)
eva = self.getOrDefault(self.evaluator)
tRatio = self.getOrDefault(self.trainRatio)
seed = self.getOrDefault(self.seed)
randCol = self.uid + "_rand"
df = dataset.select("*", rand(seed).alias(randCol))
metrics = [0.0] * numModels
condition = (df[randCol] >= tRatio)
validation = df.filter(condition)
train = df.filter(~condition)
for j in range(numModels):
model = est.fit(train, epm[j])
metric = eva.evaluate(model.transform(validation, epm[j]))
metrics[j] += metric
if eva.isLargerBetter():
bestIndex = np.argmax(metrics)
else:
bestIndex = np.argmin(metrics)
bestModel = est.fit(dataset, epm[bestIndex])
return self._copyValues(TrainValidationSplitModel(bestModel, metrics))
@since("2.0.0")
def copy(self, extra=None):
"""
Creates a copy of this instance with a randomly generated uid
and some extra params. This copies creates a deep copy of
the embedded paramMap, and copies the embedded and extra parameters over.
:param extra: Extra parameters to copy to the new instance
:return: Copy of this instance
"""
if extra is None:
extra = dict()
newTVS = Params.copy(self, extra)
if self.isSet(self.estimator):
newTVS.setEstimator(self.getEstimator().copy(extra))
# estimatorParamMaps remain the same
if self.isSet(self.evaluator):
newTVS.setEvaluator(self.getEvaluator().copy(extra))
return newTVS
class TrainValidationSplitModel(Model, ValidatorParams):
"""
.. note:: Experimental
Model from train validation split.
.. versionadded:: 2.0.0
"""
def __init__(self, bestModel, validationMetrics=[]):
super(TrainValidationSplitModel, self).__init__()
#: best model from cross validation
self.bestModel = bestModel
#: evaluated validation metrics
self.validationMetrics = validationMetrics
def _transform(self, dataset):
return self.bestModel.transform(dataset)
@since("2.0.0")
def copy(self, extra=None):
"""
Creates a copy of this instance with a randomly generated uid
and some extra params. This copies the underlying bestModel,
creates a deep copy of the embedded paramMap, and
copies the embedded and extra parameters over.
And, this creates a shallow copy of the validationMetrics.
:param extra: Extra parameters to copy to the new instance
:return: Copy of this instance
"""
if extra is None:
extra = dict()
bestModel = self.bestModel.copy(extra)
validationMetrics = list(self.validationMetrics)
return TrainValidationSplitModel(bestModel, validationMetrics)
if __name__ == "__main__":
import doctest
from pyspark.context import SparkContext
from pyspark.sql import SQLContext
globs = globals().copy()
# The small batch size here ensures that we see multiple batches,
# even in these small test examples:
sc = SparkContext("local[2]", "ml.tuning tests")
sqlContext = SQLContext(sc)
globs['sc'] = sc
globs['sqlContext'] = sqlContext
(failure_count, test_count) = doctest.testmod(globs=globs, optionflags=doctest.ELLIPSIS)
sc.stop()
if failure_count:
exit(-1)
|