2013-12-25 00:08:05 -05:00
|
|
|
#
|
|
|
|
# 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.
|
|
|
|
#
|
|
|
|
|
2014-09-19 18:01:11 -04:00
|
|
|
from math import exp
|
2018-03-08 06:38:34 -05:00
|
|
|
import sys
|
2016-04-29 01:44:14 -04:00
|
|
|
import warnings
|
2014-09-19 18:01:11 -04:00
|
|
|
|
2014-01-10 02:55:06 -05:00
|
|
|
import numpy
|
2014-09-19 18:01:11 -04:00
|
|
|
from numpy import array
|
2014-01-10 02:55:06 -05:00
|
|
|
|
2015-10-20 19:14:20 -04:00
|
|
|
from pyspark import RDD, since
|
2015-06-24 17:58:43 -04:00
|
|
|
from pyspark.streaming import DStream
|
2015-03-20 14:44:21 -04:00
|
|
|
from pyspark.mllib.common import callMLlibFunc, _py2java, _java2py
|
2015-03-31 14:32:14 -04:00
|
|
|
from pyspark.mllib.linalg import DenseVector, SparseVector, _convert_to_vector
|
2015-06-30 13:25:59 -04:00
|
|
|
from pyspark.mllib.regression import (
|
|
|
|
LabeledPoint, LinearModel, _regression_train_wrapper,
|
|
|
|
StreamingLinearAlgorithm)
|
2015-03-20 14:53:59 -04:00
|
|
|
from pyspark.mllib.util import Saveable, Loader, inherit_doc
|
2013-12-25 00:08:05 -05:00
|
|
|
|
2014-05-25 20:15:01 -04:00
|
|
|
|
[SPARK-4306] [MLlib] Python API for LogisticRegressionWithLBFGS
```
class LogisticRegressionWithLBFGS
| train(cls, data, iterations=100, initialWeights=None, corrections=10, tolerance=0.0001, regParam=0.01, intercept=False)
| Train a logistic regression model on the given data.
|
| :param data: The training data, an RDD of LabeledPoint.
| :param iterations: The number of iterations (default: 100).
| :param initialWeights: The initial weights (default: None).
| :param regParam: The regularizer parameter (default: 0.01).
| :param regType: The type of regularizer used for training
| our model.
| :Allowed values:
| - "l1" for using L1 regularization
| - "l2" for using L2 regularization
| - None for no regularization
| (default: "l2")
| :param intercept: Boolean parameter which indicates the use
| or not of the augmented representation for
| training data (i.e. whether bias features
| are activated or not).
| :param corrections: The number of corrections used in the LBFGS update (default: 10).
| :param tolerance: The convergence tolerance of iterations for L-BFGS (default: 1e-4).
|
| >>> data = [
| ... LabeledPoint(0.0, [0.0, 1.0]),
| ... LabeledPoint(1.0, [1.0, 0.0]),
| ... ]
| >>> lrm = LogisticRegressionWithLBFGS.train(sc.parallelize(data))
| >>> lrm.predict([1.0, 0.0])
| 1
| >>> lrm.predict([0.0, 1.0])
| 0
| >>> lrm.predict(sc.parallelize([[1.0, 0.0], [0.0, 1.0]])).collect()
| [1, 0]
```
Author: Davies Liu <davies@databricks.com>
Closes #3307 from davies/lbfgs and squashes the following commits:
34bd986 [Davies Liu] Merge branch 'master' of http://git-wip-us.apache.org/repos/asf/spark into lbfgs
5a945a6 [Davies Liu] address comments
941061b [Davies Liu] Merge branch 'master' of github.com:apache/spark into lbfgs
03e5543 [Davies Liu] add it to docs
ed2f9a8 [Davies Liu] add regType
76cd1b6 [Davies Liu] reorder arguments
4429a74 [Davies Liu] Update classification.py
9252783 [Davies Liu] python api for LogisticRegressionWithLBFGS
2014-11-18 18:57:33 -05:00
|
|
|
__all__ = ['LogisticRegressionModel', 'LogisticRegressionWithSGD', 'LogisticRegressionWithLBFGS',
|
2015-06-24 17:58:43 -04:00
|
|
|
'SVMModel', 'SVMWithSGD', 'NaiveBayesModel', 'NaiveBayes',
|
|
|
|
'StreamingLogisticRegressionWithSGD']
|
2014-09-03 14:49:45 -04:00
|
|
|
|
|
|
|
|
2015-03-31 14:32:14 -04:00
|
|
|
class LinearClassificationModel(LinearModel):
|
2014-11-18 13:11:13 -05:00
|
|
|
"""
|
2015-06-16 17:30:30 -04:00
|
|
|
A private abstract class representing a multiclass classification
|
|
|
|
model. The categories are represented by int values: 0, 1, 2, etc.
|
2014-11-18 13:11:13 -05:00
|
|
|
"""
|
|
|
|
def __init__(self, weights, intercept):
|
2015-03-31 14:32:14 -04:00
|
|
|
super(LinearClassificationModel, self).__init__(weights, intercept)
|
2014-11-18 13:11:13 -05:00
|
|
|
self._threshold = None
|
|
|
|
|
2015-10-20 19:14:20 -04:00
|
|
|
@since('1.4.0')
|
2014-11-18 13:11:13 -05:00
|
|
|
def setThreshold(self, value):
|
|
|
|
"""
|
2015-06-16 17:30:30 -04:00
|
|
|
Sets the threshold that separates positive predictions from
|
|
|
|
negative predictions. An example with prediction score greater
|
2016-06-06 04:35:47 -04:00
|
|
|
than or equal to this threshold is identified as a positive,
|
2015-06-16 17:30:30 -04:00
|
|
|
and negative otherwise. It is used for binary classification
|
|
|
|
only.
|
2014-11-18 13:11:13 -05:00
|
|
|
"""
|
|
|
|
self._threshold = value
|
|
|
|
|
2015-03-31 14:32:14 -04:00
|
|
|
@property
|
2015-10-20 19:14:20 -04:00
|
|
|
@since('1.4.0')
|
2015-03-31 14:32:14 -04:00
|
|
|
def threshold(self):
|
|
|
|
"""
|
2015-06-16 17:30:30 -04:00
|
|
|
Returns the threshold (if any) used for converting raw
|
|
|
|
prediction scores into 0/1 predictions. It is used for
|
|
|
|
binary classification only.
|
2015-03-31 14:32:14 -04:00
|
|
|
"""
|
|
|
|
return self._threshold
|
|
|
|
|
2015-10-20 19:14:20 -04:00
|
|
|
@since('1.4.0')
|
2014-11-18 13:11:13 -05:00
|
|
|
def clearThreshold(self):
|
|
|
|
"""
|
2015-06-16 17:30:30 -04:00
|
|
|
Clears the threshold so that `predict` will output raw
|
|
|
|
prediction scores. It is used for binary classification only.
|
2014-11-18 13:11:13 -05:00
|
|
|
"""
|
|
|
|
self._threshold = None
|
|
|
|
|
2015-10-20 19:14:20 -04:00
|
|
|
@since('1.4.0')
|
2014-11-18 13:11:13 -05:00
|
|
|
def predict(self, test):
|
|
|
|
"""
|
2015-06-16 17:30:30 -04:00
|
|
|
Predict values for a single data point or an RDD of points
|
|
|
|
using the model trained.
|
2014-11-18 13:11:13 -05:00
|
|
|
"""
|
|
|
|
raise NotImplementedError
|
|
|
|
|
|
|
|
|
2015-03-31 14:32:14 -04:00
|
|
|
class LogisticRegressionModel(LinearClassificationModel):
|
2014-08-06 15:58:24 -04:00
|
|
|
|
2015-06-16 17:30:30 -04:00
|
|
|
"""
|
|
|
|
Classification model trained using Multinomial/Binary Logistic
|
|
|
|
Regression.
|
|
|
|
|
2016-02-12 17:24:24 -05:00
|
|
|
:param weights:
|
|
|
|
Weights computed for every feature.
|
|
|
|
:param intercept:
|
|
|
|
Intercept computed for this model. (Only used in Binary Logistic
|
|
|
|
Regression. In Multinomial Logistic Regression, the intercepts will
|
|
|
|
not bea single value, so the intercepts will be part of the
|
|
|
|
weights.)
|
|
|
|
:param numFeatures:
|
|
|
|
The dimension of the features.
|
|
|
|
:param numClasses:
|
|
|
|
The number of possible outcomes for k classes classification problem
|
|
|
|
in Multinomial Logistic Regression. By default, it is binary
|
|
|
|
logistic regression so numClasses will be set to 2.
|
2013-12-25 00:08:05 -05:00
|
|
|
|
[WIP] SPARK-1430: Support sparse data in Python MLlib
This PR adds a SparseVector class in PySpark and updates all the regression, classification and clustering algorithms and models to support sparse data, similar to MLlib. I chose to add this class because SciPy is quite difficult to install in many environments (more so than NumPy), but I plan to add support for SciPy sparse vectors later too, and make the methods work transparently on objects of either type.
On the Scala side, we keep Python sparse vectors sparse and pass them to MLlib. We always return dense vectors from our models.
Some to-do items left:
- [x] Support SciPy's scipy.sparse matrix objects when SciPy is available. We can easily add a function to convert these to our own SparseVector.
- [x] MLlib currently uses a vector with one extra column on the left to represent what we call LabeledPoint in Scala. Do we really want this? It may get annoying once you deal with sparse data since you must add/subtract 1 to each feature index when training. We can remove this API in 1.0 and use tuples for labeling.
- [x] Explain how to use these in the Python MLlib docs.
CC @mengxr, @joshrosen
Author: Matei Zaharia <matei@databricks.com>
Closes #341 from mateiz/py-ml-update and squashes the following commits:
d52e763 [Matei Zaharia] Remove no-longer-needed slice code and handle review comments
ea5a25a [Matei Zaharia] Fix remaining uses of copyto() after merge
b9f97a3 [Matei Zaharia] Fix test
1e1bd0f [Matei Zaharia] Add MLlib logistic regression example in Python
88bc01f [Matei Zaharia] Clean up inheritance of LinearModel in Python, and expose its parametrs
37ab747 [Matei Zaharia] Fix some examples and docs due to changes in MLlib API
da0f27e [Matei Zaharia] Added a MLlib K-means example and updated docs to discuss sparse data
c48e85a [Matei Zaharia] Added some tests for passing lists as input, and added mllib/tests.py to run-tests script.
a07ba10 [Matei Zaharia] Fix some typos and calculation of initial weights
74eefe7 [Matei Zaharia] Added LabeledPoint class in Python
889dde8 [Matei Zaharia] Support scipy.sparse matrices in all our algorithms and models
ab244d1 [Matei Zaharia] Allow SparseVectors to be initialized using a dict
a5d6426 [Matei Zaharia] Add linalg.py to run-tests script
0e7a3d8 [Matei Zaharia] Keep vectors sparse in Java when reading LabeledPoints
eaee759 [Matei Zaharia] Update regression, classification and clustering models for sparse data
2abbb44 [Matei Zaharia] Further work to get linear models working with sparse data
154f45d [Matei Zaharia] Update docs, name some magic values
881fef7 [Matei Zaharia] Added a sparse vector in Python and made Java-Python format more compact
2014-04-15 23:33:24 -04:00
|
|
|
>>> data = [
|
2014-11-18 13:11:13 -05:00
|
|
|
... LabeledPoint(0.0, [0.0, 1.0]),
|
|
|
|
... LabeledPoint(1.0, [1.0, 0.0]),
|
[WIP] SPARK-1430: Support sparse data in Python MLlib
This PR adds a SparseVector class in PySpark and updates all the regression, classification and clustering algorithms and models to support sparse data, similar to MLlib. I chose to add this class because SciPy is quite difficult to install in many environments (more so than NumPy), but I plan to add support for SciPy sparse vectors later too, and make the methods work transparently on objects of either type.
On the Scala side, we keep Python sparse vectors sparse and pass them to MLlib. We always return dense vectors from our models.
Some to-do items left:
- [x] Support SciPy's scipy.sparse matrix objects when SciPy is available. We can easily add a function to convert these to our own SparseVector.
- [x] MLlib currently uses a vector with one extra column on the left to represent what we call LabeledPoint in Scala. Do we really want this? It may get annoying once you deal with sparse data since you must add/subtract 1 to each feature index when training. We can remove this API in 1.0 and use tuples for labeling.
- [x] Explain how to use these in the Python MLlib docs.
CC @mengxr, @joshrosen
Author: Matei Zaharia <matei@databricks.com>
Closes #341 from mateiz/py-ml-update and squashes the following commits:
d52e763 [Matei Zaharia] Remove no-longer-needed slice code and handle review comments
ea5a25a [Matei Zaharia] Fix remaining uses of copyto() after merge
b9f97a3 [Matei Zaharia] Fix test
1e1bd0f [Matei Zaharia] Add MLlib logistic regression example in Python
88bc01f [Matei Zaharia] Clean up inheritance of LinearModel in Python, and expose its parametrs
37ab747 [Matei Zaharia] Fix some examples and docs due to changes in MLlib API
da0f27e [Matei Zaharia] Added a MLlib K-means example and updated docs to discuss sparse data
c48e85a [Matei Zaharia] Added some tests for passing lists as input, and added mllib/tests.py to run-tests script.
a07ba10 [Matei Zaharia] Fix some typos and calculation of initial weights
74eefe7 [Matei Zaharia] Added LabeledPoint class in Python
889dde8 [Matei Zaharia] Support scipy.sparse matrices in all our algorithms and models
ab244d1 [Matei Zaharia] Allow SparseVectors to be initialized using a dict
a5d6426 [Matei Zaharia] Add linalg.py to run-tests script
0e7a3d8 [Matei Zaharia] Keep vectors sparse in Java when reading LabeledPoints
eaee759 [Matei Zaharia] Update regression, classification and clustering models for sparse data
2abbb44 [Matei Zaharia] Further work to get linear models working with sparse data
154f45d [Matei Zaharia] Update docs, name some magic values
881fef7 [Matei Zaharia] Added a sparse vector in Python and made Java-Python format more compact
2014-04-15 23:33:24 -04:00
|
|
|
... ]
|
2015-04-21 20:49:55 -04:00
|
|
|
>>> lrm = LogisticRegressionWithSGD.train(sc.parallelize(data), iterations=10)
|
2014-11-18 13:11:13 -05:00
|
|
|
>>> lrm.predict([1.0, 0.0])
|
|
|
|
1
|
|
|
|
>>> lrm.predict([0.0, 1.0])
|
|
|
|
0
|
|
|
|
>>> lrm.predict(sc.parallelize([[1.0, 0.0], [0.0, 1.0]])).collect()
|
|
|
|
[1, 0]
|
|
|
|
>>> lrm.clearThreshold()
|
|
|
|
>>> lrm.predict([0.0, 1.0])
|
2015-04-21 20:49:55 -04:00
|
|
|
0.279...
|
2014-11-18 13:11:13 -05:00
|
|
|
|
[WIP] SPARK-1430: Support sparse data in Python MLlib
This PR adds a SparseVector class in PySpark and updates all the regression, classification and clustering algorithms and models to support sparse data, similar to MLlib. I chose to add this class because SciPy is quite difficult to install in many environments (more so than NumPy), but I plan to add support for SciPy sparse vectors later too, and make the methods work transparently on objects of either type.
On the Scala side, we keep Python sparse vectors sparse and pass them to MLlib. We always return dense vectors from our models.
Some to-do items left:
- [x] Support SciPy's scipy.sparse matrix objects when SciPy is available. We can easily add a function to convert these to our own SparseVector.
- [x] MLlib currently uses a vector with one extra column on the left to represent what we call LabeledPoint in Scala. Do we really want this? It may get annoying once you deal with sparse data since you must add/subtract 1 to each feature index when training. We can remove this API in 1.0 and use tuples for labeling.
- [x] Explain how to use these in the Python MLlib docs.
CC @mengxr, @joshrosen
Author: Matei Zaharia <matei@databricks.com>
Closes #341 from mateiz/py-ml-update and squashes the following commits:
d52e763 [Matei Zaharia] Remove no-longer-needed slice code and handle review comments
ea5a25a [Matei Zaharia] Fix remaining uses of copyto() after merge
b9f97a3 [Matei Zaharia] Fix test
1e1bd0f [Matei Zaharia] Add MLlib logistic regression example in Python
88bc01f [Matei Zaharia] Clean up inheritance of LinearModel in Python, and expose its parametrs
37ab747 [Matei Zaharia] Fix some examples and docs due to changes in MLlib API
da0f27e [Matei Zaharia] Added a MLlib K-means example and updated docs to discuss sparse data
c48e85a [Matei Zaharia] Added some tests for passing lists as input, and added mllib/tests.py to run-tests script.
a07ba10 [Matei Zaharia] Fix some typos and calculation of initial weights
74eefe7 [Matei Zaharia] Added LabeledPoint class in Python
889dde8 [Matei Zaharia] Support scipy.sparse matrices in all our algorithms and models
ab244d1 [Matei Zaharia] Allow SparseVectors to be initialized using a dict
a5d6426 [Matei Zaharia] Add linalg.py to run-tests script
0e7a3d8 [Matei Zaharia] Keep vectors sparse in Java when reading LabeledPoints
eaee759 [Matei Zaharia] Update regression, classification and clustering models for sparse data
2abbb44 [Matei Zaharia] Further work to get linear models working with sparse data
154f45d [Matei Zaharia] Update docs, name some magic values
881fef7 [Matei Zaharia] Added a sparse vector in Python and made Java-Python format more compact
2014-04-15 23:33:24 -04:00
|
|
|
>>> sparse_data = [
|
|
|
|
... LabeledPoint(0.0, SparseVector(2, {0: 0.0})),
|
|
|
|
... LabeledPoint(1.0, SparseVector(2, {1: 1.0})),
|
2014-11-18 13:11:13 -05:00
|
|
|
... LabeledPoint(0.0, SparseVector(2, {0: 1.0})),
|
[WIP] SPARK-1430: Support sparse data in Python MLlib
This PR adds a SparseVector class in PySpark and updates all the regression, classification and clustering algorithms and models to support sparse data, similar to MLlib. I chose to add this class because SciPy is quite difficult to install in many environments (more so than NumPy), but I plan to add support for SciPy sparse vectors later too, and make the methods work transparently on objects of either type.
On the Scala side, we keep Python sparse vectors sparse and pass them to MLlib. We always return dense vectors from our models.
Some to-do items left:
- [x] Support SciPy's scipy.sparse matrix objects when SciPy is available. We can easily add a function to convert these to our own SparseVector.
- [x] MLlib currently uses a vector with one extra column on the left to represent what we call LabeledPoint in Scala. Do we really want this? It may get annoying once you deal with sparse data since you must add/subtract 1 to each feature index when training. We can remove this API in 1.0 and use tuples for labeling.
- [x] Explain how to use these in the Python MLlib docs.
CC @mengxr, @joshrosen
Author: Matei Zaharia <matei@databricks.com>
Closes #341 from mateiz/py-ml-update and squashes the following commits:
d52e763 [Matei Zaharia] Remove no-longer-needed slice code and handle review comments
ea5a25a [Matei Zaharia] Fix remaining uses of copyto() after merge
b9f97a3 [Matei Zaharia] Fix test
1e1bd0f [Matei Zaharia] Add MLlib logistic regression example in Python
88bc01f [Matei Zaharia] Clean up inheritance of LinearModel in Python, and expose its parametrs
37ab747 [Matei Zaharia] Fix some examples and docs due to changes in MLlib API
da0f27e [Matei Zaharia] Added a MLlib K-means example and updated docs to discuss sparse data
c48e85a [Matei Zaharia] Added some tests for passing lists as input, and added mllib/tests.py to run-tests script.
a07ba10 [Matei Zaharia] Fix some typos and calculation of initial weights
74eefe7 [Matei Zaharia] Added LabeledPoint class in Python
889dde8 [Matei Zaharia] Support scipy.sparse matrices in all our algorithms and models
ab244d1 [Matei Zaharia] Allow SparseVectors to be initialized using a dict
a5d6426 [Matei Zaharia] Add linalg.py to run-tests script
0e7a3d8 [Matei Zaharia] Keep vectors sparse in Java when reading LabeledPoints
eaee759 [Matei Zaharia] Update regression, classification and clustering models for sparse data
2abbb44 [Matei Zaharia] Further work to get linear models working with sparse data
154f45d [Matei Zaharia] Update docs, name some magic values
881fef7 [Matei Zaharia] Added a sparse vector in Python and made Java-Python format more compact
2014-04-15 23:33:24 -04:00
|
|
|
... LabeledPoint(1.0, SparseVector(2, {1: 2.0}))
|
|
|
|
... ]
|
2015-04-21 20:49:55 -04:00
|
|
|
>>> lrm = LogisticRegressionWithSGD.train(sc.parallelize(sparse_data), iterations=10)
|
2014-11-18 13:11:13 -05:00
|
|
|
>>> lrm.predict(array([0.0, 1.0]))
|
|
|
|
1
|
|
|
|
>>> lrm.predict(array([1.0, 0.0]))
|
|
|
|
0
|
|
|
|
>>> lrm.predict(SparseVector(2, {1: 1.0}))
|
|
|
|
1
|
|
|
|
>>> lrm.predict(SparseVector(2, {0: 1.0}))
|
|
|
|
0
|
2015-03-20 14:44:21 -04:00
|
|
|
>>> import os, tempfile
|
|
|
|
>>> path = tempfile.mkdtemp()
|
|
|
|
>>> lrm.save(sc, path)
|
|
|
|
>>> sameModel = LogisticRegressionModel.load(sc, path)
|
|
|
|
>>> sameModel.predict(array([0.0, 1.0]))
|
|
|
|
1
|
|
|
|
>>> sameModel.predict(SparseVector(2, {0: 1.0}))
|
|
|
|
0
|
2015-06-22 14:53:11 -04:00
|
|
|
>>> from shutil import rmtree
|
2015-03-20 14:44:21 -04:00
|
|
|
>>> try:
|
2015-06-22 14:53:11 -04:00
|
|
|
... rmtree(path)
|
2015-03-20 14:44:21 -04:00
|
|
|
... except:
|
|
|
|
... pass
|
2015-03-31 14:32:14 -04:00
|
|
|
>>> multi_class_data = [
|
|
|
|
... LabeledPoint(0.0, [0.0, 1.0, 0.0]),
|
|
|
|
... LabeledPoint(1.0, [1.0, 0.0, 0.0]),
|
|
|
|
... LabeledPoint(2.0, [0.0, 0.0, 1.0])
|
|
|
|
... ]
|
2015-04-21 20:49:55 -04:00
|
|
|
>>> data = sc.parallelize(multi_class_data)
|
|
|
|
>>> mcm = LogisticRegressionWithLBFGS.train(data, iterations=10, numClasses=3)
|
2015-03-31 14:32:14 -04:00
|
|
|
>>> mcm.predict([0.0, 0.5, 0.0])
|
|
|
|
0
|
|
|
|
>>> mcm.predict([0.8, 0.0, 0.0])
|
|
|
|
1
|
|
|
|
>>> mcm.predict([0.0, 0.0, 0.3])
|
|
|
|
2
|
2015-10-20 19:14:20 -04:00
|
|
|
|
|
|
|
.. versionadded:: 0.9.0
|
2013-12-25 00:08:05 -05:00
|
|
|
"""
|
2015-03-31 14:32:14 -04:00
|
|
|
def __init__(self, weights, intercept, numFeatures, numClasses):
|
2014-11-18 13:11:13 -05:00
|
|
|
super(LogisticRegressionModel, self).__init__(weights, intercept)
|
2015-03-31 14:32:14 -04:00
|
|
|
self._numFeatures = int(numFeatures)
|
|
|
|
self._numClasses = int(numClasses)
|
2014-11-18 13:11:13 -05:00
|
|
|
self._threshold = 0.5
|
2015-03-31 14:32:14 -04:00
|
|
|
if self._numClasses == 2:
|
|
|
|
self._dataWithBiasSize = None
|
|
|
|
self._weightsMatrix = None
|
|
|
|
else:
|
2017-05-24 10:55:38 -04:00
|
|
|
self._dataWithBiasSize = self._coeff.size // (self._numClasses - 1)
|
2015-03-31 14:32:14 -04:00
|
|
|
self._weightsMatrix = self._coeff.toArray().reshape(self._numClasses - 1,
|
|
|
|
self._dataWithBiasSize)
|
|
|
|
|
|
|
|
@property
|
2015-10-20 19:14:20 -04:00
|
|
|
@since('1.4.0')
|
2015-03-31 14:32:14 -04:00
|
|
|
def numFeatures(self):
|
2015-10-20 19:14:20 -04:00
|
|
|
"""
|
|
|
|
Dimension of the features.
|
|
|
|
"""
|
2015-03-31 14:32:14 -04:00
|
|
|
return self._numFeatures
|
|
|
|
|
|
|
|
@property
|
2015-10-20 19:14:20 -04:00
|
|
|
@since('1.4.0')
|
2015-03-31 14:32:14 -04:00
|
|
|
def numClasses(self):
|
2015-10-20 19:14:20 -04:00
|
|
|
"""
|
2016-02-12 17:24:24 -05:00
|
|
|
Number of possible outcomes for k classes classification problem
|
|
|
|
in Multinomial Logistic Regression.
|
2015-10-20 19:14:20 -04:00
|
|
|
"""
|
2015-03-31 14:32:14 -04:00
|
|
|
return self._numClasses
|
2014-08-06 15:58:24 -04:00
|
|
|
|
2015-10-20 19:14:20 -04:00
|
|
|
@since('0.9.0')
|
2013-12-25 00:08:05 -05:00
|
|
|
def predict(self, x):
|
2014-11-18 13:11:13 -05:00
|
|
|
"""
|
2015-06-16 17:30:30 -04:00
|
|
|
Predict values for a single data point or an RDD of points
|
|
|
|
using the model trained.
|
2014-11-18 13:11:13 -05:00
|
|
|
"""
|
|
|
|
if isinstance(x, RDD):
|
|
|
|
return x.map(lambda v: self.predict(v))
|
|
|
|
|
2014-11-11 01:26:16 -05:00
|
|
|
x = _convert_to_vector(x)
|
2015-03-31 14:32:14 -04:00
|
|
|
if self.numClasses == 2:
|
|
|
|
margin = self.weights.dot(x) + self._intercept
|
|
|
|
if margin > 0:
|
|
|
|
prob = 1 / (1 + exp(-margin))
|
|
|
|
else:
|
|
|
|
exp_margin = exp(margin)
|
|
|
|
prob = exp_margin / (1 + exp_margin)
|
|
|
|
if self._threshold is None:
|
|
|
|
return prob
|
|
|
|
else:
|
|
|
|
return 1 if prob > self._threshold else 0
|
2014-07-20 21:40:36 -04:00
|
|
|
else:
|
2015-03-31 14:32:14 -04:00
|
|
|
best_class = 0
|
|
|
|
max_margin = 0.0
|
|
|
|
if x.size + 1 == self._dataWithBiasSize:
|
|
|
|
for i in range(0, self._numClasses - 1):
|
|
|
|
margin = x.dot(self._weightsMatrix[i][0:x.size]) + \
|
|
|
|
self._weightsMatrix[i][x.size]
|
|
|
|
if margin > max_margin:
|
|
|
|
max_margin = margin
|
|
|
|
best_class = i + 1
|
|
|
|
else:
|
|
|
|
for i in range(0, self._numClasses - 1):
|
|
|
|
margin = x.dot(self._weightsMatrix[i])
|
|
|
|
if margin > max_margin:
|
|
|
|
max_margin = margin
|
|
|
|
best_class = i + 1
|
|
|
|
return best_class
|
2013-12-25 00:08:05 -05:00
|
|
|
|
2015-10-20 19:14:20 -04:00
|
|
|
@since('1.4.0')
|
2015-03-20 14:44:21 -04:00
|
|
|
def save(self, sc, path):
|
2015-10-20 19:14:20 -04:00
|
|
|
"""
|
|
|
|
Save this model to the given path.
|
|
|
|
"""
|
2015-03-20 14:44:21 -04:00
|
|
|
java_model = sc._jvm.org.apache.spark.mllib.classification.LogisticRegressionModel(
|
2015-03-31 14:32:14 -04:00
|
|
|
_py2java(sc, self._coeff), self.intercept, self.numFeatures, self.numClasses)
|
2015-03-20 14:44:21 -04:00
|
|
|
java_model.save(sc._jsc.sc(), path)
|
|
|
|
|
|
|
|
@classmethod
|
2015-10-20 19:14:20 -04:00
|
|
|
@since('1.4.0')
|
2015-03-20 14:44:21 -04:00
|
|
|
def load(cls, sc, path):
|
2015-10-20 19:14:20 -04:00
|
|
|
"""
|
|
|
|
Load a model from the given path.
|
|
|
|
"""
|
2015-03-20 14:44:21 -04:00
|
|
|
java_model = sc._jvm.org.apache.spark.mllib.classification.LogisticRegressionModel.load(
|
|
|
|
sc._jsc.sc(), path)
|
|
|
|
weights = _java2py(sc, java_model.weights())
|
|
|
|
intercept = java_model.intercept()
|
2015-03-31 14:32:14 -04:00
|
|
|
numFeatures = java_model.numFeatures()
|
|
|
|
numClasses = java_model.numClasses()
|
2015-03-20 14:44:21 -04:00
|
|
|
threshold = java_model.getThreshold().get()
|
2015-03-31 14:32:14 -04:00
|
|
|
model = LogisticRegressionModel(weights, intercept, numFeatures, numClasses)
|
2015-03-20 14:44:21 -04:00
|
|
|
model.setThreshold(threshold)
|
|
|
|
return model
|
|
|
|
|
2018-06-28 15:40:39 -04:00
|
|
|
def __repr__(self):
|
|
|
|
return self._call_java("toString")
|
|
|
|
|
[WIP] SPARK-1430: Support sparse data in Python MLlib
This PR adds a SparseVector class in PySpark and updates all the regression, classification and clustering algorithms and models to support sparse data, similar to MLlib. I chose to add this class because SciPy is quite difficult to install in many environments (more so than NumPy), but I plan to add support for SciPy sparse vectors later too, and make the methods work transparently on objects of either type.
On the Scala side, we keep Python sparse vectors sparse and pass them to MLlib. We always return dense vectors from our models.
Some to-do items left:
- [x] Support SciPy's scipy.sparse matrix objects when SciPy is available. We can easily add a function to convert these to our own SparseVector.
- [x] MLlib currently uses a vector with one extra column on the left to represent what we call LabeledPoint in Scala. Do we really want this? It may get annoying once you deal with sparse data since you must add/subtract 1 to each feature index when training. We can remove this API in 1.0 and use tuples for labeling.
- [x] Explain how to use these in the Python MLlib docs.
CC @mengxr, @joshrosen
Author: Matei Zaharia <matei@databricks.com>
Closes #341 from mateiz/py-ml-update and squashes the following commits:
d52e763 [Matei Zaharia] Remove no-longer-needed slice code and handle review comments
ea5a25a [Matei Zaharia] Fix remaining uses of copyto() after merge
b9f97a3 [Matei Zaharia] Fix test
1e1bd0f [Matei Zaharia] Add MLlib logistic regression example in Python
88bc01f [Matei Zaharia] Clean up inheritance of LinearModel in Python, and expose its parametrs
37ab747 [Matei Zaharia] Fix some examples and docs due to changes in MLlib API
da0f27e [Matei Zaharia] Added a MLlib K-means example and updated docs to discuss sparse data
c48e85a [Matei Zaharia] Added some tests for passing lists as input, and added mllib/tests.py to run-tests script.
a07ba10 [Matei Zaharia] Fix some typos and calculation of initial weights
74eefe7 [Matei Zaharia] Added LabeledPoint class in Python
889dde8 [Matei Zaharia] Support scipy.sparse matrices in all our algorithms and models
ab244d1 [Matei Zaharia] Allow SparseVectors to be initialized using a dict
a5d6426 [Matei Zaharia] Add linalg.py to run-tests script
0e7a3d8 [Matei Zaharia] Keep vectors sparse in Java when reading LabeledPoints
eaee759 [Matei Zaharia] Update regression, classification and clustering models for sparse data
2abbb44 [Matei Zaharia] Further work to get linear models working with sparse data
154f45d [Matei Zaharia] Update docs, name some magic values
881fef7 [Matei Zaharia] Added a sparse vector in Python and made Java-Python format more compact
2014-04-15 23:33:24 -04:00
|
|
|
|
2013-12-25 00:08:05 -05:00
|
|
|
class LogisticRegressionWithSGD(object):
|
2015-10-20 19:14:20 -04:00
|
|
|
"""
|
|
|
|
.. versionadded:: 0.9.0
|
2016-04-29 01:44:14 -04:00
|
|
|
.. note:: Deprecated in 2.0.0. Use ml.classification.LogisticRegression or
|
|
|
|
LogisticRegressionWithLBFGS.
|
2015-10-20 19:14:20 -04:00
|
|
|
"""
|
2013-12-25 00:08:05 -05:00
|
|
|
@classmethod
|
2015-10-20 19:14:20 -04:00
|
|
|
@since('0.9.0')
|
2014-08-05 19:30:32 -04:00
|
|
|
def train(cls, data, iterations=100, step=1.0, miniBatchFraction=1.0,
|
2015-03-31 14:32:14 -04:00
|
|
|
initialWeights=None, regParam=0.01, regType="l2", intercept=False,
|
2015-09-14 15:08:52 -04:00
|
|
|
validateData=True, convergenceTol=0.001):
|
2014-08-05 19:30:32 -04:00
|
|
|
"""
|
|
|
|
Train a logistic regression model on the given data.
|
|
|
|
|
2016-02-12 17:24:24 -05:00
|
|
|
:param data:
|
|
|
|
The training data, an RDD of LabeledPoint.
|
|
|
|
:param iterations:
|
|
|
|
The number of iterations.
|
|
|
|
(default: 100)
|
|
|
|
:param step:
|
|
|
|
The step parameter used in SGD.
|
|
|
|
(default: 1.0)
|
|
|
|
:param miniBatchFraction:
|
|
|
|
Fraction of data to be used for each SGD iteration.
|
|
|
|
(default: 1.0)
|
|
|
|
:param initialWeights:
|
|
|
|
The initial weights.
|
|
|
|
(default: None)
|
|
|
|
:param regParam:
|
|
|
|
The regularizer parameter.
|
|
|
|
(default: 0.01)
|
|
|
|
:param regType:
|
|
|
|
The type of regularizer used for training our model.
|
2016-02-29 08:52:41 -05:00
|
|
|
Supported values:
|
2016-02-12 17:24:24 -05:00
|
|
|
|
|
|
|
- "l1" for using L1 regularization
|
|
|
|
- "l2" for using L2 regularization (default)
|
|
|
|
- None for no regularization
|
|
|
|
:param intercept:
|
|
|
|
Boolean parameter which indicates the use or not of the
|
|
|
|
augmented representation for training data (i.e., whether bias
|
|
|
|
features are activated or not).
|
|
|
|
(default: False)
|
|
|
|
:param validateData:
|
|
|
|
Boolean parameter which indicates if the algorithm should
|
|
|
|
validate data before training.
|
|
|
|
(default: True)
|
|
|
|
:param convergenceTol:
|
|
|
|
A condition which decides iteration termination.
|
|
|
|
(default: 0.001)
|
2014-08-05 19:30:32 -04:00
|
|
|
"""
|
2016-04-29 01:44:14 -04:00
|
|
|
warnings.warn(
|
|
|
|
"Deprecated in 2.0.0. Use ml.classification.LogisticRegression or "
|
[SPARK-22313][PYTHON] Mark/print deprecation warnings as DeprecationWarning for deprecated APIs
## What changes were proposed in this pull request?
This PR proposes to mark the existing warnings as `DeprecationWarning` and print out warnings for deprecated functions.
This could be actually useful for Spark app developers. I use (old) PyCharm and this IDE can detect this specific `DeprecationWarning` in some cases:
**Before**
<img src="https://user-images.githubusercontent.com/6477701/31762664-df68d9f8-b4f6-11e7-8773-f0468f70a2cc.png" height="45" />
**After**
<img src="https://user-images.githubusercontent.com/6477701/31762662-de4d6868-b4f6-11e7-98dc-3c8446a0c28a.png" height="70" />
For console usage, `DeprecationWarning` is usually disabled (see https://docs.python.org/2/library/warnings.html#warning-categories and https://docs.python.org/3/library/warnings.html#warning-categories):
```
>>> import warnings
>>> filter(lambda f: f[2] == DeprecationWarning, warnings.filters)
[('ignore', <_sre.SRE_Pattern object at 0x10ba58c00>, <type 'exceptions.DeprecationWarning'>, <_sre.SRE_Pattern object at 0x10bb04138>, 0), ('ignore', None, <type 'exceptions.DeprecationWarning'>, None, 0)]
```
so, it won't actually mess up the terminal much unless it is intended.
If this is intendedly enabled, it'd should as below:
```
>>> import warnings
>>> warnings.simplefilter('always', DeprecationWarning)
>>>
>>> from pyspark.sql import functions
>>> functions.approxCountDistinct("a")
.../spark/python/pyspark/sql/functions.py:232: DeprecationWarning: Deprecated in 2.1, use approx_count_distinct instead.
"Deprecated in 2.1, use approx_count_distinct instead.", DeprecationWarning)
...
```
These instances were found by:
```
cd python/pyspark
grep -r "Deprecated" .
grep -r "deprecated" .
grep -r "deprecate" .
```
## How was this patch tested?
Manually tested.
Author: hyukjinkwon <gurwls223@gmail.com>
Closes #19535 from HyukjinKwon/deprecated-warning.
2017-10-23 23:44:47 -04:00
|
|
|
"LogisticRegressionWithLBFGS.", DeprecationWarning)
|
2016-04-29 01:44:14 -04:00
|
|
|
|
2014-10-31 01:25:18 -04:00
|
|
|
def train(rdd, i):
|
2014-11-13 16:54:16 -05:00
|
|
|
return callMLlibFunc("trainLogisticRegressionModelWithSGD", rdd, int(iterations),
|
|
|
|
float(step), float(miniBatchFraction), i, float(regParam), regType,
|
2015-09-14 15:08:52 -04:00
|
|
|
bool(intercept), bool(validateData), float(convergenceTol))
|
2014-09-19 18:01:11 -04:00
|
|
|
|
2014-10-31 01:25:18 -04:00
|
|
|
return _regression_train_wrapper(train, LogisticRegressionModel, data, initialWeights)
|
2014-05-25 20:15:01 -04:00
|
|
|
|
2013-12-25 00:08:05 -05:00
|
|
|
|
[SPARK-4306] [MLlib] Python API for LogisticRegressionWithLBFGS
```
class LogisticRegressionWithLBFGS
| train(cls, data, iterations=100, initialWeights=None, corrections=10, tolerance=0.0001, regParam=0.01, intercept=False)
| Train a logistic regression model on the given data.
|
| :param data: The training data, an RDD of LabeledPoint.
| :param iterations: The number of iterations (default: 100).
| :param initialWeights: The initial weights (default: None).
| :param regParam: The regularizer parameter (default: 0.01).
| :param regType: The type of regularizer used for training
| our model.
| :Allowed values:
| - "l1" for using L1 regularization
| - "l2" for using L2 regularization
| - None for no regularization
| (default: "l2")
| :param intercept: Boolean parameter which indicates the use
| or not of the augmented representation for
| training data (i.e. whether bias features
| are activated or not).
| :param corrections: The number of corrections used in the LBFGS update (default: 10).
| :param tolerance: The convergence tolerance of iterations for L-BFGS (default: 1e-4).
|
| >>> data = [
| ... LabeledPoint(0.0, [0.0, 1.0]),
| ... LabeledPoint(1.0, [1.0, 0.0]),
| ... ]
| >>> lrm = LogisticRegressionWithLBFGS.train(sc.parallelize(data))
| >>> lrm.predict([1.0, 0.0])
| 1
| >>> lrm.predict([0.0, 1.0])
| 0
| >>> lrm.predict(sc.parallelize([[1.0, 0.0], [0.0, 1.0]])).collect()
| [1, 0]
```
Author: Davies Liu <davies@databricks.com>
Closes #3307 from davies/lbfgs and squashes the following commits:
34bd986 [Davies Liu] Merge branch 'master' of http://git-wip-us.apache.org/repos/asf/spark into lbfgs
5a945a6 [Davies Liu] address comments
941061b [Davies Liu] Merge branch 'master' of github.com:apache/spark into lbfgs
03e5543 [Davies Liu] add it to docs
ed2f9a8 [Davies Liu] add regType
76cd1b6 [Davies Liu] reorder arguments
4429a74 [Davies Liu] Update classification.py
9252783 [Davies Liu] python api for LogisticRegressionWithLBFGS
2014-11-18 18:57:33 -05:00
|
|
|
class LogisticRegressionWithLBFGS(object):
|
2015-10-20 19:14:20 -04:00
|
|
|
"""
|
|
|
|
.. versionadded:: 1.2.0
|
|
|
|
"""
|
[SPARK-4306] [MLlib] Python API for LogisticRegressionWithLBFGS
```
class LogisticRegressionWithLBFGS
| train(cls, data, iterations=100, initialWeights=None, corrections=10, tolerance=0.0001, regParam=0.01, intercept=False)
| Train a logistic regression model on the given data.
|
| :param data: The training data, an RDD of LabeledPoint.
| :param iterations: The number of iterations (default: 100).
| :param initialWeights: The initial weights (default: None).
| :param regParam: The regularizer parameter (default: 0.01).
| :param regType: The type of regularizer used for training
| our model.
| :Allowed values:
| - "l1" for using L1 regularization
| - "l2" for using L2 regularization
| - None for no regularization
| (default: "l2")
| :param intercept: Boolean parameter which indicates the use
| or not of the augmented representation for
| training data (i.e. whether bias features
| are activated or not).
| :param corrections: The number of corrections used in the LBFGS update (default: 10).
| :param tolerance: The convergence tolerance of iterations for L-BFGS (default: 1e-4).
|
| >>> data = [
| ... LabeledPoint(0.0, [0.0, 1.0]),
| ... LabeledPoint(1.0, [1.0, 0.0]),
| ... ]
| >>> lrm = LogisticRegressionWithLBFGS.train(sc.parallelize(data))
| >>> lrm.predict([1.0, 0.0])
| 1
| >>> lrm.predict([0.0, 1.0])
| 0
| >>> lrm.predict(sc.parallelize([[1.0, 0.0], [0.0, 1.0]])).collect()
| [1, 0]
```
Author: Davies Liu <davies@databricks.com>
Closes #3307 from davies/lbfgs and squashes the following commits:
34bd986 [Davies Liu] Merge branch 'master' of http://git-wip-us.apache.org/repos/asf/spark into lbfgs
5a945a6 [Davies Liu] address comments
941061b [Davies Liu] Merge branch 'master' of github.com:apache/spark into lbfgs
03e5543 [Davies Liu] add it to docs
ed2f9a8 [Davies Liu] add regType
76cd1b6 [Davies Liu] reorder arguments
4429a74 [Davies Liu] Update classification.py
9252783 [Davies Liu] python api for LogisticRegressionWithLBFGS
2014-11-18 18:57:33 -05:00
|
|
|
@classmethod
|
2015-10-20 19:14:20 -04:00
|
|
|
@since('1.2.0')
|
2016-02-29 03:55:51 -05:00
|
|
|
def train(cls, data, iterations=100, initialWeights=None, regParam=0.0, regType="l2",
|
2016-02-23 02:37:09 -05:00
|
|
|
intercept=False, corrections=10, tolerance=1e-6, validateData=True, numClasses=2):
|
[SPARK-4306] [MLlib] Python API for LogisticRegressionWithLBFGS
```
class LogisticRegressionWithLBFGS
| train(cls, data, iterations=100, initialWeights=None, corrections=10, tolerance=0.0001, regParam=0.01, intercept=False)
| Train a logistic regression model on the given data.
|
| :param data: The training data, an RDD of LabeledPoint.
| :param iterations: The number of iterations (default: 100).
| :param initialWeights: The initial weights (default: None).
| :param regParam: The regularizer parameter (default: 0.01).
| :param regType: The type of regularizer used for training
| our model.
| :Allowed values:
| - "l1" for using L1 regularization
| - "l2" for using L2 regularization
| - None for no regularization
| (default: "l2")
| :param intercept: Boolean parameter which indicates the use
| or not of the augmented representation for
| training data (i.e. whether bias features
| are activated or not).
| :param corrections: The number of corrections used in the LBFGS update (default: 10).
| :param tolerance: The convergence tolerance of iterations for L-BFGS (default: 1e-4).
|
| >>> data = [
| ... LabeledPoint(0.0, [0.0, 1.0]),
| ... LabeledPoint(1.0, [1.0, 0.0]),
| ... ]
| >>> lrm = LogisticRegressionWithLBFGS.train(sc.parallelize(data))
| >>> lrm.predict([1.0, 0.0])
| 1
| >>> lrm.predict([0.0, 1.0])
| 0
| >>> lrm.predict(sc.parallelize([[1.0, 0.0], [0.0, 1.0]])).collect()
| [1, 0]
```
Author: Davies Liu <davies@databricks.com>
Closes #3307 from davies/lbfgs and squashes the following commits:
34bd986 [Davies Liu] Merge branch 'master' of http://git-wip-us.apache.org/repos/asf/spark into lbfgs
5a945a6 [Davies Liu] address comments
941061b [Davies Liu] Merge branch 'master' of github.com:apache/spark into lbfgs
03e5543 [Davies Liu] add it to docs
ed2f9a8 [Davies Liu] add regType
76cd1b6 [Davies Liu] reorder arguments
4429a74 [Davies Liu] Update classification.py
9252783 [Davies Liu] python api for LogisticRegressionWithLBFGS
2014-11-18 18:57:33 -05:00
|
|
|
"""
|
|
|
|
Train a logistic regression model on the given data.
|
|
|
|
|
2016-02-12 17:24:24 -05:00
|
|
|
:param data:
|
|
|
|
The training data, an RDD of LabeledPoint.
|
|
|
|
:param iterations:
|
|
|
|
The number of iterations.
|
|
|
|
(default: 100)
|
|
|
|
:param initialWeights:
|
|
|
|
The initial weights.
|
|
|
|
(default: None)
|
|
|
|
:param regParam:
|
|
|
|
The regularizer parameter.
|
2016-02-29 03:55:51 -05:00
|
|
|
(default: 0.0)
|
2016-02-12 17:24:24 -05:00
|
|
|
:param regType:
|
|
|
|
The type of regularizer used for training our model.
|
2016-02-29 08:52:41 -05:00
|
|
|
Supported values:
|
2016-02-12 17:24:24 -05:00
|
|
|
|
|
|
|
- "l1" for using L1 regularization
|
|
|
|
- "l2" for using L2 regularization (default)
|
|
|
|
- None for no regularization
|
|
|
|
:param intercept:
|
|
|
|
Boolean parameter which indicates the use or not of the
|
|
|
|
augmented representation for training data (i.e., whether bias
|
|
|
|
features are activated or not).
|
|
|
|
(default: False)
|
|
|
|
:param corrections:
|
|
|
|
The number of corrections used in the LBFGS update.
|
2016-02-29 03:55:51 -05:00
|
|
|
If a known updater is used for binary classification,
|
|
|
|
it calls the ml implementation and this parameter will
|
|
|
|
have no effect. (default: 10)
|
2016-02-12 17:24:24 -05:00
|
|
|
:param tolerance:
|
|
|
|
The convergence tolerance of iterations for L-BFGS.
|
2016-02-23 02:37:09 -05:00
|
|
|
(default: 1e-6)
|
2016-02-12 17:24:24 -05:00
|
|
|
:param validateData:
|
|
|
|
Boolean parameter which indicates if the algorithm should
|
|
|
|
validate data before training.
|
|
|
|
(default: True)
|
|
|
|
:param numClasses:
|
|
|
|
The number of classes (i.e., outcomes) a label can take in
|
|
|
|
Multinomial Logistic Regression.
|
|
|
|
(default: 2)
|
[SPARK-4306] [MLlib] Python API for LogisticRegressionWithLBFGS
```
class LogisticRegressionWithLBFGS
| train(cls, data, iterations=100, initialWeights=None, corrections=10, tolerance=0.0001, regParam=0.01, intercept=False)
| Train a logistic regression model on the given data.
|
| :param data: The training data, an RDD of LabeledPoint.
| :param iterations: The number of iterations (default: 100).
| :param initialWeights: The initial weights (default: None).
| :param regParam: The regularizer parameter (default: 0.01).
| :param regType: The type of regularizer used for training
| our model.
| :Allowed values:
| - "l1" for using L1 regularization
| - "l2" for using L2 regularization
| - None for no regularization
| (default: "l2")
| :param intercept: Boolean parameter which indicates the use
| or not of the augmented representation for
| training data (i.e. whether bias features
| are activated or not).
| :param corrections: The number of corrections used in the LBFGS update (default: 10).
| :param tolerance: The convergence tolerance of iterations for L-BFGS (default: 1e-4).
|
| >>> data = [
| ... LabeledPoint(0.0, [0.0, 1.0]),
| ... LabeledPoint(1.0, [1.0, 0.0]),
| ... ]
| >>> lrm = LogisticRegressionWithLBFGS.train(sc.parallelize(data))
| >>> lrm.predict([1.0, 0.0])
| 1
| >>> lrm.predict([0.0, 1.0])
| 0
| >>> lrm.predict(sc.parallelize([[1.0, 0.0], [0.0, 1.0]])).collect()
| [1, 0]
```
Author: Davies Liu <davies@databricks.com>
Closes #3307 from davies/lbfgs and squashes the following commits:
34bd986 [Davies Liu] Merge branch 'master' of http://git-wip-us.apache.org/repos/asf/spark into lbfgs
5a945a6 [Davies Liu] address comments
941061b [Davies Liu] Merge branch 'master' of github.com:apache/spark into lbfgs
03e5543 [Davies Liu] add it to docs
ed2f9a8 [Davies Liu] add regType
76cd1b6 [Davies Liu] reorder arguments
4429a74 [Davies Liu] Update classification.py
9252783 [Davies Liu] python api for LogisticRegressionWithLBFGS
2014-11-18 18:57:33 -05:00
|
|
|
|
|
|
|
>>> data = [
|
|
|
|
... LabeledPoint(0.0, [0.0, 1.0]),
|
|
|
|
... LabeledPoint(1.0, [1.0, 0.0]),
|
|
|
|
... ]
|
2015-04-21 20:49:55 -04:00
|
|
|
>>> lrm = LogisticRegressionWithLBFGS.train(sc.parallelize(data), iterations=10)
|
[SPARK-4306] [MLlib] Python API for LogisticRegressionWithLBFGS
```
class LogisticRegressionWithLBFGS
| train(cls, data, iterations=100, initialWeights=None, corrections=10, tolerance=0.0001, regParam=0.01, intercept=False)
| Train a logistic regression model on the given data.
|
| :param data: The training data, an RDD of LabeledPoint.
| :param iterations: The number of iterations (default: 100).
| :param initialWeights: The initial weights (default: None).
| :param regParam: The regularizer parameter (default: 0.01).
| :param regType: The type of regularizer used for training
| our model.
| :Allowed values:
| - "l1" for using L1 regularization
| - "l2" for using L2 regularization
| - None for no regularization
| (default: "l2")
| :param intercept: Boolean parameter which indicates the use
| or not of the augmented representation for
| training data (i.e. whether bias features
| are activated or not).
| :param corrections: The number of corrections used in the LBFGS update (default: 10).
| :param tolerance: The convergence tolerance of iterations for L-BFGS (default: 1e-4).
|
| >>> data = [
| ... LabeledPoint(0.0, [0.0, 1.0]),
| ... LabeledPoint(1.0, [1.0, 0.0]),
| ... ]
| >>> lrm = LogisticRegressionWithLBFGS.train(sc.parallelize(data))
| >>> lrm.predict([1.0, 0.0])
| 1
| >>> lrm.predict([0.0, 1.0])
| 0
| >>> lrm.predict(sc.parallelize([[1.0, 0.0], [0.0, 1.0]])).collect()
| [1, 0]
```
Author: Davies Liu <davies@databricks.com>
Closes #3307 from davies/lbfgs and squashes the following commits:
34bd986 [Davies Liu] Merge branch 'master' of http://git-wip-us.apache.org/repos/asf/spark into lbfgs
5a945a6 [Davies Liu] address comments
941061b [Davies Liu] Merge branch 'master' of github.com:apache/spark into lbfgs
03e5543 [Davies Liu] add it to docs
ed2f9a8 [Davies Liu] add regType
76cd1b6 [Davies Liu] reorder arguments
4429a74 [Davies Liu] Update classification.py
9252783 [Davies Liu] python api for LogisticRegressionWithLBFGS
2014-11-18 18:57:33 -05:00
|
|
|
>>> lrm.predict([1.0, 0.0])
|
|
|
|
1
|
|
|
|
>>> lrm.predict([0.0, 1.0])
|
|
|
|
0
|
|
|
|
"""
|
|
|
|
def train(rdd, i):
|
|
|
|
return callMLlibFunc("trainLogisticRegressionModelWithLBFGS", rdd, int(iterations), i,
|
2015-03-02 13:17:24 -05:00
|
|
|
float(regParam), regType, bool(intercept), int(corrections),
|
2015-03-31 14:32:14 -04:00
|
|
|
float(tolerance), bool(validateData), int(numClasses))
|
|
|
|
|
|
|
|
if initialWeights is None:
|
|
|
|
if numClasses == 2:
|
|
|
|
initialWeights = [0.0] * len(data.first().features)
|
|
|
|
else:
|
|
|
|
if intercept:
|
|
|
|
initialWeights = [0.0] * (len(data.first().features) + 1) * (numClasses - 1)
|
|
|
|
else:
|
|
|
|
initialWeights = [0.0] * len(data.first().features) * (numClasses - 1)
|
[SPARK-4306] [MLlib] Python API for LogisticRegressionWithLBFGS
```
class LogisticRegressionWithLBFGS
| train(cls, data, iterations=100, initialWeights=None, corrections=10, tolerance=0.0001, regParam=0.01, intercept=False)
| Train a logistic regression model on the given data.
|
| :param data: The training data, an RDD of LabeledPoint.
| :param iterations: The number of iterations (default: 100).
| :param initialWeights: The initial weights (default: None).
| :param regParam: The regularizer parameter (default: 0.01).
| :param regType: The type of regularizer used for training
| our model.
| :Allowed values:
| - "l1" for using L1 regularization
| - "l2" for using L2 regularization
| - None for no regularization
| (default: "l2")
| :param intercept: Boolean parameter which indicates the use
| or not of the augmented representation for
| training data (i.e. whether bias features
| are activated or not).
| :param corrections: The number of corrections used in the LBFGS update (default: 10).
| :param tolerance: The convergence tolerance of iterations for L-BFGS (default: 1e-4).
|
| >>> data = [
| ... LabeledPoint(0.0, [0.0, 1.0]),
| ... LabeledPoint(1.0, [1.0, 0.0]),
| ... ]
| >>> lrm = LogisticRegressionWithLBFGS.train(sc.parallelize(data))
| >>> lrm.predict([1.0, 0.0])
| 1
| >>> lrm.predict([0.0, 1.0])
| 0
| >>> lrm.predict(sc.parallelize([[1.0, 0.0], [0.0, 1.0]])).collect()
| [1, 0]
```
Author: Davies Liu <davies@databricks.com>
Closes #3307 from davies/lbfgs and squashes the following commits:
34bd986 [Davies Liu] Merge branch 'master' of http://git-wip-us.apache.org/repos/asf/spark into lbfgs
5a945a6 [Davies Liu] address comments
941061b [Davies Liu] Merge branch 'master' of github.com:apache/spark into lbfgs
03e5543 [Davies Liu] add it to docs
ed2f9a8 [Davies Liu] add regType
76cd1b6 [Davies Liu] reorder arguments
4429a74 [Davies Liu] Update classification.py
9252783 [Davies Liu] python api for LogisticRegressionWithLBFGS
2014-11-18 18:57:33 -05:00
|
|
|
return _regression_train_wrapper(train, LogisticRegressionModel, data, initialWeights)
|
|
|
|
|
|
|
|
|
2015-03-31 14:32:14 -04:00
|
|
|
class SVMModel(LinearClassificationModel):
|
2014-08-06 15:58:24 -04:00
|
|
|
|
2015-06-16 17:30:30 -04:00
|
|
|
"""
|
|
|
|
Model for Support Vector Machines (SVMs).
|
|
|
|
|
2016-02-12 17:24:24 -05:00
|
|
|
:param weights:
|
|
|
|
Weights computed for every feature.
|
|
|
|
:param intercept:
|
|
|
|
Intercept computed for this model.
|
2013-12-25 00:08:05 -05:00
|
|
|
|
[WIP] SPARK-1430: Support sparse data in Python MLlib
This PR adds a SparseVector class in PySpark and updates all the regression, classification and clustering algorithms and models to support sparse data, similar to MLlib. I chose to add this class because SciPy is quite difficult to install in many environments (more so than NumPy), but I plan to add support for SciPy sparse vectors later too, and make the methods work transparently on objects of either type.
On the Scala side, we keep Python sparse vectors sparse and pass them to MLlib. We always return dense vectors from our models.
Some to-do items left:
- [x] Support SciPy's scipy.sparse matrix objects when SciPy is available. We can easily add a function to convert these to our own SparseVector.
- [x] MLlib currently uses a vector with one extra column on the left to represent what we call LabeledPoint in Scala. Do we really want this? It may get annoying once you deal with sparse data since you must add/subtract 1 to each feature index when training. We can remove this API in 1.0 and use tuples for labeling.
- [x] Explain how to use these in the Python MLlib docs.
CC @mengxr, @joshrosen
Author: Matei Zaharia <matei@databricks.com>
Closes #341 from mateiz/py-ml-update and squashes the following commits:
d52e763 [Matei Zaharia] Remove no-longer-needed slice code and handle review comments
ea5a25a [Matei Zaharia] Fix remaining uses of copyto() after merge
b9f97a3 [Matei Zaharia] Fix test
1e1bd0f [Matei Zaharia] Add MLlib logistic regression example in Python
88bc01f [Matei Zaharia] Clean up inheritance of LinearModel in Python, and expose its parametrs
37ab747 [Matei Zaharia] Fix some examples and docs due to changes in MLlib API
da0f27e [Matei Zaharia] Added a MLlib K-means example and updated docs to discuss sparse data
c48e85a [Matei Zaharia] Added some tests for passing lists as input, and added mllib/tests.py to run-tests script.
a07ba10 [Matei Zaharia] Fix some typos and calculation of initial weights
74eefe7 [Matei Zaharia] Added LabeledPoint class in Python
889dde8 [Matei Zaharia] Support scipy.sparse matrices in all our algorithms and models
ab244d1 [Matei Zaharia] Allow SparseVectors to be initialized using a dict
a5d6426 [Matei Zaharia] Add linalg.py to run-tests script
0e7a3d8 [Matei Zaharia] Keep vectors sparse in Java when reading LabeledPoints
eaee759 [Matei Zaharia] Update regression, classification and clustering models for sparse data
2abbb44 [Matei Zaharia] Further work to get linear models working with sparse data
154f45d [Matei Zaharia] Update docs, name some magic values
881fef7 [Matei Zaharia] Added a sparse vector in Python and made Java-Python format more compact
2014-04-15 23:33:24 -04:00
|
|
|
>>> data = [
|
|
|
|
... LabeledPoint(0.0, [0.0]),
|
|
|
|
... LabeledPoint(1.0, [1.0]),
|
|
|
|
... LabeledPoint(1.0, [2.0]),
|
|
|
|
... LabeledPoint(1.0, [3.0])
|
|
|
|
... ]
|
2015-04-21 20:49:55 -04:00
|
|
|
>>> svm = SVMWithSGD.train(sc.parallelize(data), iterations=10)
|
2014-11-18 13:11:13 -05:00
|
|
|
>>> svm.predict([1.0])
|
|
|
|
1
|
|
|
|
>>> svm.predict(sc.parallelize([[1.0]])).collect()
|
|
|
|
[1]
|
|
|
|
>>> svm.clearThreshold()
|
|
|
|
>>> svm.predict(array([1.0]))
|
2015-04-21 20:49:55 -04:00
|
|
|
1.44...
|
2014-11-18 13:11:13 -05:00
|
|
|
|
[WIP] SPARK-1430: Support sparse data in Python MLlib
This PR adds a SparseVector class in PySpark and updates all the regression, classification and clustering algorithms and models to support sparse data, similar to MLlib. I chose to add this class because SciPy is quite difficult to install in many environments (more so than NumPy), but I plan to add support for SciPy sparse vectors later too, and make the methods work transparently on objects of either type.
On the Scala side, we keep Python sparse vectors sparse and pass them to MLlib. We always return dense vectors from our models.
Some to-do items left:
- [x] Support SciPy's scipy.sparse matrix objects when SciPy is available. We can easily add a function to convert these to our own SparseVector.
- [x] MLlib currently uses a vector with one extra column on the left to represent what we call LabeledPoint in Scala. Do we really want this? It may get annoying once you deal with sparse data since you must add/subtract 1 to each feature index when training. We can remove this API in 1.0 and use tuples for labeling.
- [x] Explain how to use these in the Python MLlib docs.
CC @mengxr, @joshrosen
Author: Matei Zaharia <matei@databricks.com>
Closes #341 from mateiz/py-ml-update and squashes the following commits:
d52e763 [Matei Zaharia] Remove no-longer-needed slice code and handle review comments
ea5a25a [Matei Zaharia] Fix remaining uses of copyto() after merge
b9f97a3 [Matei Zaharia] Fix test
1e1bd0f [Matei Zaharia] Add MLlib logistic regression example in Python
88bc01f [Matei Zaharia] Clean up inheritance of LinearModel in Python, and expose its parametrs
37ab747 [Matei Zaharia] Fix some examples and docs due to changes in MLlib API
da0f27e [Matei Zaharia] Added a MLlib K-means example and updated docs to discuss sparse data
c48e85a [Matei Zaharia] Added some tests for passing lists as input, and added mllib/tests.py to run-tests script.
a07ba10 [Matei Zaharia] Fix some typos and calculation of initial weights
74eefe7 [Matei Zaharia] Added LabeledPoint class in Python
889dde8 [Matei Zaharia] Support scipy.sparse matrices in all our algorithms and models
ab244d1 [Matei Zaharia] Allow SparseVectors to be initialized using a dict
a5d6426 [Matei Zaharia] Add linalg.py to run-tests script
0e7a3d8 [Matei Zaharia] Keep vectors sparse in Java when reading LabeledPoints
eaee759 [Matei Zaharia] Update regression, classification and clustering models for sparse data
2abbb44 [Matei Zaharia] Further work to get linear models working with sparse data
154f45d [Matei Zaharia] Update docs, name some magic values
881fef7 [Matei Zaharia] Added a sparse vector in Python and made Java-Python format more compact
2014-04-15 23:33:24 -04:00
|
|
|
>>> sparse_data = [
|
2014-05-05 21:32:54 -04:00
|
|
|
... LabeledPoint(0.0, SparseVector(2, {0: -1.0})),
|
[WIP] SPARK-1430: Support sparse data in Python MLlib
This PR adds a SparseVector class in PySpark and updates all the regression, classification and clustering algorithms and models to support sparse data, similar to MLlib. I chose to add this class because SciPy is quite difficult to install in many environments (more so than NumPy), but I plan to add support for SciPy sparse vectors later too, and make the methods work transparently on objects of either type.
On the Scala side, we keep Python sparse vectors sparse and pass them to MLlib. We always return dense vectors from our models.
Some to-do items left:
- [x] Support SciPy's scipy.sparse matrix objects when SciPy is available. We can easily add a function to convert these to our own SparseVector.
- [x] MLlib currently uses a vector with one extra column on the left to represent what we call LabeledPoint in Scala. Do we really want this? It may get annoying once you deal with sparse data since you must add/subtract 1 to each feature index when training. We can remove this API in 1.0 and use tuples for labeling.
- [x] Explain how to use these in the Python MLlib docs.
CC @mengxr, @joshrosen
Author: Matei Zaharia <matei@databricks.com>
Closes #341 from mateiz/py-ml-update and squashes the following commits:
d52e763 [Matei Zaharia] Remove no-longer-needed slice code and handle review comments
ea5a25a [Matei Zaharia] Fix remaining uses of copyto() after merge
b9f97a3 [Matei Zaharia] Fix test
1e1bd0f [Matei Zaharia] Add MLlib logistic regression example in Python
88bc01f [Matei Zaharia] Clean up inheritance of LinearModel in Python, and expose its parametrs
37ab747 [Matei Zaharia] Fix some examples and docs due to changes in MLlib API
da0f27e [Matei Zaharia] Added a MLlib K-means example and updated docs to discuss sparse data
c48e85a [Matei Zaharia] Added some tests for passing lists as input, and added mllib/tests.py to run-tests script.
a07ba10 [Matei Zaharia] Fix some typos and calculation of initial weights
74eefe7 [Matei Zaharia] Added LabeledPoint class in Python
889dde8 [Matei Zaharia] Support scipy.sparse matrices in all our algorithms and models
ab244d1 [Matei Zaharia] Allow SparseVectors to be initialized using a dict
a5d6426 [Matei Zaharia] Add linalg.py to run-tests script
0e7a3d8 [Matei Zaharia] Keep vectors sparse in Java when reading LabeledPoints
eaee759 [Matei Zaharia] Update regression, classification and clustering models for sparse data
2abbb44 [Matei Zaharia] Further work to get linear models working with sparse data
154f45d [Matei Zaharia] Update docs, name some magic values
881fef7 [Matei Zaharia] Added a sparse vector in Python and made Java-Python format more compact
2014-04-15 23:33:24 -04:00
|
|
|
... LabeledPoint(1.0, SparseVector(2, {1: 1.0})),
|
|
|
|
... LabeledPoint(0.0, SparseVector(2, {0: 0.0})),
|
|
|
|
... LabeledPoint(1.0, SparseVector(2, {1: 2.0}))
|
|
|
|
... ]
|
2015-04-21 20:49:55 -04:00
|
|
|
>>> svm = SVMWithSGD.train(sc.parallelize(sparse_data), iterations=10)
|
2014-11-18 13:11:13 -05:00
|
|
|
>>> svm.predict(SparseVector(2, {1: 1.0}))
|
|
|
|
1
|
|
|
|
>>> svm.predict(SparseVector(2, {0: -1.0}))
|
|
|
|
0
|
2015-03-20 14:44:21 -04:00
|
|
|
>>> import os, tempfile
|
|
|
|
>>> path = tempfile.mkdtemp()
|
|
|
|
>>> svm.save(sc, path)
|
|
|
|
>>> sameModel = SVMModel.load(sc, path)
|
|
|
|
>>> sameModel.predict(SparseVector(2, {1: 1.0}))
|
|
|
|
1
|
|
|
|
>>> sameModel.predict(SparseVector(2, {0: -1.0}))
|
|
|
|
0
|
2015-06-22 14:53:11 -04:00
|
|
|
>>> from shutil import rmtree
|
2015-03-20 14:44:21 -04:00
|
|
|
>>> try:
|
2015-06-22 14:53:11 -04:00
|
|
|
... rmtree(path)
|
2015-03-20 14:44:21 -04:00
|
|
|
... except:
|
|
|
|
... pass
|
2015-10-20 19:14:20 -04:00
|
|
|
|
|
|
|
.. versionadded:: 0.9.0
|
2013-12-25 00:08:05 -05:00
|
|
|
"""
|
2014-11-18 13:11:13 -05:00
|
|
|
def __init__(self, weights, intercept):
|
|
|
|
super(SVMModel, self).__init__(weights, intercept)
|
|
|
|
self._threshold = 0.0
|
2014-08-06 15:58:24 -04:00
|
|
|
|
2015-10-20 19:14:20 -04:00
|
|
|
@since('0.9.0')
|
2013-12-25 00:08:05 -05:00
|
|
|
def predict(self, x):
|
2014-11-18 13:11:13 -05:00
|
|
|
"""
|
2015-06-16 17:30:30 -04:00
|
|
|
Predict values for a single data point or an RDD of points
|
|
|
|
using the model trained.
|
2014-11-18 13:11:13 -05:00
|
|
|
"""
|
|
|
|
if isinstance(x, RDD):
|
|
|
|
return x.map(lambda v: self.predict(v))
|
|
|
|
|
2014-11-11 01:26:16 -05:00
|
|
|
x = _convert_to_vector(x)
|
2014-09-19 18:01:11 -04:00
|
|
|
margin = self.weights.dot(x) + self.intercept
|
2014-11-18 13:11:13 -05:00
|
|
|
if self._threshold is None:
|
|
|
|
return margin
|
|
|
|
else:
|
|
|
|
return 1 if margin > self._threshold else 0
|
2013-12-25 00:08:05 -05:00
|
|
|
|
2015-10-20 19:14:20 -04:00
|
|
|
@since('1.4.0')
|
2015-03-20 14:44:21 -04:00
|
|
|
def save(self, sc, path):
|
2015-10-20 19:14:20 -04:00
|
|
|
"""
|
|
|
|
Save this model to the given path.
|
|
|
|
"""
|
2015-03-20 14:44:21 -04:00
|
|
|
java_model = sc._jvm.org.apache.spark.mllib.classification.SVMModel(
|
|
|
|
_py2java(sc, self._coeff), self.intercept)
|
|
|
|
java_model.save(sc._jsc.sc(), path)
|
|
|
|
|
|
|
|
@classmethod
|
2015-10-20 19:14:20 -04:00
|
|
|
@since('1.4.0')
|
2015-03-20 14:44:21 -04:00
|
|
|
def load(cls, sc, path):
|
2015-10-20 19:14:20 -04:00
|
|
|
"""
|
|
|
|
Load a model from the given path.
|
|
|
|
"""
|
2015-03-20 14:44:21 -04:00
|
|
|
java_model = sc._jvm.org.apache.spark.mllib.classification.SVMModel.load(
|
|
|
|
sc._jsc.sc(), path)
|
|
|
|
weights = _java2py(sc, java_model.weights())
|
|
|
|
intercept = java_model.intercept()
|
|
|
|
threshold = java_model.getThreshold().get()
|
|
|
|
model = SVMModel(weights, intercept)
|
|
|
|
model.setThreshold(threshold)
|
|
|
|
return model
|
|
|
|
|
2014-05-25 20:15:01 -04:00
|
|
|
|
2013-12-25 00:08:05 -05:00
|
|
|
class SVMWithSGD(object):
|
2015-10-20 19:14:20 -04:00
|
|
|
"""
|
|
|
|
.. versionadded:: 0.9.0
|
|
|
|
"""
|
2014-08-06 15:58:24 -04:00
|
|
|
|
2013-12-25 00:08:05 -05:00
|
|
|
@classmethod
|
2015-10-20 19:14:20 -04:00
|
|
|
@since('0.9.0')
|
2014-11-13 16:54:16 -05:00
|
|
|
def train(cls, data, iterations=100, step=1.0, regParam=0.01,
|
2015-03-31 14:32:14 -04:00
|
|
|
miniBatchFraction=1.0, initialWeights=None, regType="l2",
|
2015-09-14 15:08:52 -04:00
|
|
|
intercept=False, validateData=True, convergenceTol=0.001):
|
2014-08-05 19:30:32 -04:00
|
|
|
"""
|
|
|
|
Train a support vector machine on the given data.
|
|
|
|
|
2016-02-12 17:24:24 -05:00
|
|
|
:param data:
|
|
|
|
The training data, an RDD of LabeledPoint.
|
|
|
|
:param iterations:
|
|
|
|
The number of iterations.
|
|
|
|
(default: 100)
|
|
|
|
:param step:
|
|
|
|
The step parameter used in SGD.
|
|
|
|
(default: 1.0)
|
|
|
|
:param regParam:
|
|
|
|
The regularizer parameter.
|
|
|
|
(default: 0.01)
|
|
|
|
:param miniBatchFraction:
|
|
|
|
Fraction of data to be used for each SGD iteration.
|
|
|
|
(default: 1.0)
|
|
|
|
:param initialWeights:
|
|
|
|
The initial weights.
|
|
|
|
(default: None)
|
|
|
|
:param regType:
|
|
|
|
The type of regularizer used for training our model.
|
|
|
|
Allowed values:
|
|
|
|
|
|
|
|
- "l1" for using L1 regularization
|
|
|
|
- "l2" for using L2 regularization (default)
|
|
|
|
- None for no regularization
|
|
|
|
:param intercept:
|
|
|
|
Boolean parameter which indicates the use or not of the
|
|
|
|
augmented representation for training data (i.e. whether bias
|
|
|
|
features are activated or not).
|
|
|
|
(default: False)
|
|
|
|
:param validateData:
|
|
|
|
Boolean parameter which indicates if the algorithm should
|
|
|
|
validate data before training.
|
|
|
|
(default: True)
|
|
|
|
:param convergenceTol:
|
|
|
|
A condition which decides iteration termination.
|
|
|
|
(default: 0.001)
|
2014-08-05 19:30:32 -04:00
|
|
|
"""
|
2014-10-31 01:25:18 -04:00
|
|
|
def train(rdd, i):
|
2014-11-13 16:54:16 -05:00
|
|
|
return callMLlibFunc("trainSVMModelWithSGD", rdd, int(iterations), float(step),
|
|
|
|
float(regParam), float(miniBatchFraction), i, regType,
|
2015-09-14 15:08:52 -04:00
|
|
|
bool(intercept), bool(validateData), float(convergenceTol))
|
2014-09-19 18:01:11 -04:00
|
|
|
|
2014-10-31 01:25:18 -04:00
|
|
|
return _regression_train_wrapper(train, SVMModel, data, initialWeights)
|
2014-05-25 20:15:01 -04:00
|
|
|
|
2013-12-25 00:08:05 -05:00
|
|
|
|
2015-03-20 14:53:59 -04:00
|
|
|
@inherit_doc
|
|
|
|
class NaiveBayesModel(Saveable, Loader):
|
2014-08-06 15:58:24 -04:00
|
|
|
|
2014-01-10 02:55:06 -05:00
|
|
|
"""
|
|
|
|
Model for Naive Bayes classifiers.
|
|
|
|
|
2016-02-12 17:24:24 -05:00
|
|
|
:param labels:
|
|
|
|
List of labels.
|
|
|
|
:param pi:
|
|
|
|
Log of class priors, whose dimension is C, number of labels.
|
|
|
|
:param theta:
|
|
|
|
Log of class conditional probabilities, whose dimension is C-by-D,
|
|
|
|
where D is number of features.
|
2014-01-10 02:55:06 -05:00
|
|
|
|
[WIP] SPARK-1430: Support sparse data in Python MLlib
This PR adds a SparseVector class in PySpark and updates all the regression, classification and clustering algorithms and models to support sparse data, similar to MLlib. I chose to add this class because SciPy is quite difficult to install in many environments (more so than NumPy), but I plan to add support for SciPy sparse vectors later too, and make the methods work transparently on objects of either type.
On the Scala side, we keep Python sparse vectors sparse and pass them to MLlib. We always return dense vectors from our models.
Some to-do items left:
- [x] Support SciPy's scipy.sparse matrix objects when SciPy is available. We can easily add a function to convert these to our own SparseVector.
- [x] MLlib currently uses a vector with one extra column on the left to represent what we call LabeledPoint in Scala. Do we really want this? It may get annoying once you deal with sparse data since you must add/subtract 1 to each feature index when training. We can remove this API in 1.0 and use tuples for labeling.
- [x] Explain how to use these in the Python MLlib docs.
CC @mengxr, @joshrosen
Author: Matei Zaharia <matei@databricks.com>
Closes #341 from mateiz/py-ml-update and squashes the following commits:
d52e763 [Matei Zaharia] Remove no-longer-needed slice code and handle review comments
ea5a25a [Matei Zaharia] Fix remaining uses of copyto() after merge
b9f97a3 [Matei Zaharia] Fix test
1e1bd0f [Matei Zaharia] Add MLlib logistic regression example in Python
88bc01f [Matei Zaharia] Clean up inheritance of LinearModel in Python, and expose its parametrs
37ab747 [Matei Zaharia] Fix some examples and docs due to changes in MLlib API
da0f27e [Matei Zaharia] Added a MLlib K-means example and updated docs to discuss sparse data
c48e85a [Matei Zaharia] Added some tests for passing lists as input, and added mllib/tests.py to run-tests script.
a07ba10 [Matei Zaharia] Fix some typos and calculation of initial weights
74eefe7 [Matei Zaharia] Added LabeledPoint class in Python
889dde8 [Matei Zaharia] Support scipy.sparse matrices in all our algorithms and models
ab244d1 [Matei Zaharia] Allow SparseVectors to be initialized using a dict
a5d6426 [Matei Zaharia] Add linalg.py to run-tests script
0e7a3d8 [Matei Zaharia] Keep vectors sparse in Java when reading LabeledPoints
eaee759 [Matei Zaharia] Update regression, classification and clustering models for sparse data
2abbb44 [Matei Zaharia] Further work to get linear models working with sparse data
154f45d [Matei Zaharia] Update docs, name some magic values
881fef7 [Matei Zaharia] Added a sparse vector in Python and made Java-Python format more compact
2014-04-15 23:33:24 -04:00
|
|
|
>>> data = [
|
|
|
|
... LabeledPoint(0.0, [0.0, 0.0]),
|
|
|
|
... LabeledPoint(0.0, [0.0, 1.0]),
|
|
|
|
... LabeledPoint(1.0, [1.0, 0.0]),
|
|
|
|
... ]
|
2014-01-10 02:55:06 -05:00
|
|
|
>>> model = NaiveBayes.train(sc.parallelize(data))
|
|
|
|
>>> model.predict(array([0.0, 1.0]))
|
2014-04-02 17:01:12 -04:00
|
|
|
0.0
|
2014-01-10 02:55:06 -05:00
|
|
|
>>> model.predict(array([1.0, 0.0]))
|
2014-04-02 17:01:12 -04:00
|
|
|
1.0
|
2014-11-18 13:11:13 -05:00
|
|
|
>>> model.predict(sc.parallelize([[1.0, 0.0]])).collect()
|
|
|
|
[1.0]
|
[WIP] SPARK-1430: Support sparse data in Python MLlib
This PR adds a SparseVector class in PySpark and updates all the regression, classification and clustering algorithms and models to support sparse data, similar to MLlib. I chose to add this class because SciPy is quite difficult to install in many environments (more so than NumPy), but I plan to add support for SciPy sparse vectors later too, and make the methods work transparently on objects of either type.
On the Scala side, we keep Python sparse vectors sparse and pass them to MLlib. We always return dense vectors from our models.
Some to-do items left:
- [x] Support SciPy's scipy.sparse matrix objects when SciPy is available. We can easily add a function to convert these to our own SparseVector.
- [x] MLlib currently uses a vector with one extra column on the left to represent what we call LabeledPoint in Scala. Do we really want this? It may get annoying once you deal with sparse data since you must add/subtract 1 to each feature index when training. We can remove this API in 1.0 and use tuples for labeling.
- [x] Explain how to use these in the Python MLlib docs.
CC @mengxr, @joshrosen
Author: Matei Zaharia <matei@databricks.com>
Closes #341 from mateiz/py-ml-update and squashes the following commits:
d52e763 [Matei Zaharia] Remove no-longer-needed slice code and handle review comments
ea5a25a [Matei Zaharia] Fix remaining uses of copyto() after merge
b9f97a3 [Matei Zaharia] Fix test
1e1bd0f [Matei Zaharia] Add MLlib logistic regression example in Python
88bc01f [Matei Zaharia] Clean up inheritance of LinearModel in Python, and expose its parametrs
37ab747 [Matei Zaharia] Fix some examples and docs due to changes in MLlib API
da0f27e [Matei Zaharia] Added a MLlib K-means example and updated docs to discuss sparse data
c48e85a [Matei Zaharia] Added some tests for passing lists as input, and added mllib/tests.py to run-tests script.
a07ba10 [Matei Zaharia] Fix some typos and calculation of initial weights
74eefe7 [Matei Zaharia] Added LabeledPoint class in Python
889dde8 [Matei Zaharia] Support scipy.sparse matrices in all our algorithms and models
ab244d1 [Matei Zaharia] Allow SparseVectors to be initialized using a dict
a5d6426 [Matei Zaharia] Add linalg.py to run-tests script
0e7a3d8 [Matei Zaharia] Keep vectors sparse in Java when reading LabeledPoints
eaee759 [Matei Zaharia] Update regression, classification and clustering models for sparse data
2abbb44 [Matei Zaharia] Further work to get linear models working with sparse data
154f45d [Matei Zaharia] Update docs, name some magic values
881fef7 [Matei Zaharia] Added a sparse vector in Python and made Java-Python format more compact
2014-04-15 23:33:24 -04:00
|
|
|
>>> sparse_data = [
|
|
|
|
... LabeledPoint(0.0, SparseVector(2, {1: 0.0})),
|
|
|
|
... LabeledPoint(0.0, SparseVector(2, {1: 1.0})),
|
|
|
|
... LabeledPoint(1.0, SparseVector(2, {0: 1.0}))
|
|
|
|
... ]
|
|
|
|
>>> model = NaiveBayes.train(sc.parallelize(sparse_data))
|
|
|
|
>>> model.predict(SparseVector(2, {1: 1.0}))
|
|
|
|
0.0
|
|
|
|
>>> model.predict(SparseVector(2, {0: 1.0}))
|
|
|
|
1.0
|
2015-03-20 14:53:59 -04:00
|
|
|
>>> import os, tempfile
|
|
|
|
>>> path = tempfile.mkdtemp()
|
|
|
|
>>> model.save(sc, path)
|
|
|
|
>>> sameModel = NaiveBayesModel.load(sc, path)
|
|
|
|
>>> sameModel.predict(SparseVector(2, {0: 1.0})) == model.predict(SparseVector(2, {0: 1.0}))
|
|
|
|
True
|
2015-06-22 14:53:11 -04:00
|
|
|
>>> from shutil import rmtree
|
2015-03-20 14:53:59 -04:00
|
|
|
>>> try:
|
2015-06-22 14:53:11 -04:00
|
|
|
... rmtree(path)
|
2015-03-20 14:53:59 -04:00
|
|
|
... except OSError:
|
|
|
|
... pass
|
2014-01-10 02:55:06 -05:00
|
|
|
|
2015-10-20 19:14:20 -04:00
|
|
|
.. versionadded:: 0.9.0
|
|
|
|
"""
|
2014-04-02 17:01:12 -04:00
|
|
|
def __init__(self, labels, pi, theta):
|
|
|
|
self.labels = labels
|
2014-01-10 02:55:06 -05:00
|
|
|
self.pi = pi
|
|
|
|
self.theta = theta
|
|
|
|
|
2015-10-20 19:14:20 -04:00
|
|
|
@since('0.9.0')
|
2014-01-10 02:55:06 -05:00
|
|
|
def predict(self, x):
|
2015-06-16 17:30:30 -04:00
|
|
|
"""
|
|
|
|
Return the most likely class for a data vector
|
|
|
|
or an RDD of vectors
|
|
|
|
"""
|
2014-11-18 13:11:13 -05:00
|
|
|
if isinstance(x, RDD):
|
|
|
|
return x.map(lambda v: self.predict(v))
|
2014-09-19 18:01:11 -04:00
|
|
|
x = _convert_to_vector(x)
|
|
|
|
return self.labels[numpy.argmax(self.pi + x.dot(self.theta.transpose()))]
|
2014-01-10 02:55:06 -05:00
|
|
|
|
2015-03-20 14:53:59 -04:00
|
|
|
def save(self, sc, path):
|
2015-10-20 19:14:20 -04:00
|
|
|
"""
|
|
|
|
Save this model to the given path.
|
|
|
|
"""
|
2015-03-20 14:53:59 -04:00
|
|
|
java_labels = _py2java(sc, self.labels.tolist())
|
|
|
|
java_pi = _py2java(sc, self.pi.tolist())
|
|
|
|
java_theta = _py2java(sc, self.theta.tolist())
|
|
|
|
java_model = sc._jvm.org.apache.spark.mllib.classification.NaiveBayesModel(
|
|
|
|
java_labels, java_pi, java_theta)
|
|
|
|
java_model.save(sc._jsc.sc(), path)
|
|
|
|
|
|
|
|
@classmethod
|
2015-10-20 19:14:20 -04:00
|
|
|
@since('1.4.0')
|
2015-03-20 14:53:59 -04:00
|
|
|
def load(cls, sc, path):
|
2015-10-20 19:14:20 -04:00
|
|
|
"""
|
|
|
|
Load a model from the given path.
|
|
|
|
"""
|
2015-03-20 14:53:59 -04:00
|
|
|
java_model = sc._jvm.org.apache.spark.mllib.classification.NaiveBayesModel.load(
|
|
|
|
sc._jsc.sc(), path)
|
2015-04-16 19:20:57 -04:00
|
|
|
# Can not unpickle array.array from Pyrolite in Python3 with "bytes"
|
|
|
|
py_labels = _java2py(sc, java_model.labels(), "latin1")
|
|
|
|
py_pi = _java2py(sc, java_model.pi(), "latin1")
|
|
|
|
py_theta = _java2py(sc, java_model.theta(), "latin1")
|
2015-03-20 14:53:59 -04:00
|
|
|
return NaiveBayesModel(py_labels, py_pi, numpy.array(py_theta))
|
|
|
|
|
2014-05-25 20:15:01 -04:00
|
|
|
|
2014-01-10 02:55:06 -05:00
|
|
|
class NaiveBayes(object):
|
2015-10-20 19:14:20 -04:00
|
|
|
"""
|
|
|
|
.. versionadded:: 0.9.0
|
|
|
|
"""
|
2014-08-06 15:58:24 -04:00
|
|
|
|
2014-01-10 02:55:06 -05:00
|
|
|
@classmethod
|
2015-10-20 19:14:20 -04:00
|
|
|
@since('0.9.0')
|
2014-01-10 02:55:06 -05:00
|
|
|
def train(cls, data, lambda_=1.0):
|
|
|
|
"""
|
2015-06-16 17:30:30 -04:00
|
|
|
Train a Naive Bayes model given an RDD of (label, features)
|
|
|
|
vectors.
|
2014-01-10 02:55:06 -05:00
|
|
|
|
2015-06-16 17:30:30 -04:00
|
|
|
This is the Multinomial NB (U{http://tinyurl.com/lsdw6p}) which
|
|
|
|
can handle all kinds of discrete data. For example, by
|
|
|
|
converting documents into TF-IDF vectors, it can be used for
|
|
|
|
document classification. By making every vector a 0-1 vector,
|
|
|
|
it can also be used as Bernoulli NB (U{http://tinyurl.com/p7c96j6}).
|
|
|
|
The input feature values must be nonnegative.
|
2014-01-10 02:55:06 -05:00
|
|
|
|
2016-02-12 17:24:24 -05:00
|
|
|
:param data:
|
|
|
|
RDD of LabeledPoint.
|
|
|
|
:param lambda_:
|
|
|
|
The smoothing parameter.
|
|
|
|
(default: 1.0)
|
2014-01-10 02:55:06 -05:00
|
|
|
"""
|
2014-11-11 01:26:16 -05:00
|
|
|
first = data.first()
|
|
|
|
if not isinstance(first, LabeledPoint):
|
|
|
|
raise ValueError("`data` should be an RDD of LabeledPoint")
|
2015-06-25 11:13:17 -04:00
|
|
|
labels, pi, theta = callMLlibFunc("trainNaiveBayesModel", data, lambda_)
|
2014-09-19 18:01:11 -04:00
|
|
|
return NaiveBayesModel(labels.toArray(), pi.toArray(), numpy.array(theta))
|
2014-01-10 02:55:06 -05:00
|
|
|
|
|
|
|
|
2015-06-24 17:58:43 -04:00
|
|
|
@inherit_doc
|
|
|
|
class StreamingLogisticRegressionWithSGD(StreamingLinearAlgorithm):
|
|
|
|
"""
|
2016-02-12 17:24:24 -05:00
|
|
|
Train or predict a logistic regression model on streaming data.
|
|
|
|
Training uses Stochastic Gradient Descent to update the model based on
|
|
|
|
each new batch of incoming data from a DStream.
|
2015-11-23 20:11:51 -05:00
|
|
|
|
|
|
|
Each batch of data is assumed to be an RDD of LabeledPoints.
|
|
|
|
The number of data points per batch can vary, but the number
|
|
|
|
of features must be constant. An initial weight
|
|
|
|
vector must be provided.
|
|
|
|
|
|
|
|
:param stepSize:
|
|
|
|
Step size for each iteration of gradient descent.
|
|
|
|
(default: 0.1)
|
|
|
|
:param numIterations:
|
|
|
|
Number of iterations run for each batch of data.
|
|
|
|
(default: 50)
|
|
|
|
:param miniBatchFraction:
|
|
|
|
Fraction of each batch of data to use for updates.
|
|
|
|
(default: 1.0)
|
|
|
|
:param regParam:
|
|
|
|
L2 Regularization parameter.
|
|
|
|
(default: 0.0)
|
|
|
|
:param convergenceTol:
|
|
|
|
Value used to determine when to terminate iterations.
|
|
|
|
(default: 0.001)
|
2015-10-20 19:14:20 -04:00
|
|
|
|
|
|
|
.. versionadded:: 1.5.0
|
2015-06-24 17:58:43 -04:00
|
|
|
"""
|
2015-11-23 20:11:51 -05:00
|
|
|
def __init__(self, stepSize=0.1, numIterations=50, miniBatchFraction=1.0, regParam=0.0,
|
2015-09-14 15:08:52 -04:00
|
|
|
convergenceTol=0.001):
|
2015-06-24 17:58:43 -04:00
|
|
|
self.stepSize = stepSize
|
|
|
|
self.numIterations = numIterations
|
|
|
|
self.regParam = regParam
|
|
|
|
self.miniBatchFraction = miniBatchFraction
|
2015-09-14 15:08:52 -04:00
|
|
|
self.convergenceTol = convergenceTol
|
2015-06-24 17:58:43 -04:00
|
|
|
self._model = None
|
|
|
|
super(StreamingLogisticRegressionWithSGD, self).__init__(
|
|
|
|
model=self._model)
|
|
|
|
|
2015-10-20 19:14:20 -04:00
|
|
|
@since('1.5.0')
|
2015-06-24 17:58:43 -04:00
|
|
|
def setInitialWeights(self, initialWeights):
|
|
|
|
"""
|
|
|
|
Set the initial value of weights.
|
|
|
|
|
|
|
|
This must be set before running trainOn and predictOn.
|
|
|
|
"""
|
|
|
|
initialWeights = _convert_to_vector(initialWeights)
|
|
|
|
|
|
|
|
# LogisticRegressionWithSGD does only binary classification.
|
|
|
|
self._model = LogisticRegressionModel(
|
|
|
|
initialWeights, 0, initialWeights.size, 2)
|
|
|
|
return self
|
|
|
|
|
2015-10-20 19:14:20 -04:00
|
|
|
@since('1.5.0')
|
2015-06-24 17:58:43 -04:00
|
|
|
def trainOn(self, dstream):
|
|
|
|
"""Train the model on the incoming dstream."""
|
|
|
|
self._validate(dstream)
|
|
|
|
|
|
|
|
def update(rdd):
|
|
|
|
# LogisticRegressionWithSGD.train raises an error for an empty RDD.
|
|
|
|
if not rdd.isEmpty():
|
|
|
|
self._model = LogisticRegressionWithSGD.train(
|
|
|
|
rdd, self.numIterations, self.stepSize,
|
2015-10-09 01:21:07 -04:00
|
|
|
self.miniBatchFraction, self._model.weights,
|
|
|
|
regParam=self.regParam, convergenceTol=self.convergenceTol)
|
2015-06-24 17:58:43 -04:00
|
|
|
|
|
|
|
dstream.foreachRDD(update)
|
|
|
|
|
|
|
|
|
2013-12-25 00:08:05 -05:00
|
|
|
def _test():
|
|
|
|
import doctest
|
2016-05-23 21:14:48 -04:00
|
|
|
from pyspark.sql import SparkSession
|
2014-11-18 13:11:13 -05:00
|
|
|
import pyspark.mllib.classification
|
|
|
|
globs = pyspark.mllib.classification.__dict__.copy()
|
2016-05-23 21:14:48 -04:00
|
|
|
spark = SparkSession.builder\
|
|
|
|
.master("local[4]")\
|
|
|
|
.appName("mllib.classification tests")\
|
|
|
|
.getOrCreate()
|
|
|
|
globs['sc'] = spark.sparkContext
|
2014-05-25 20:15:01 -04:00
|
|
|
(failure_count, test_count) = doctest.testmod(globs=globs, optionflags=doctest.ELLIPSIS)
|
2016-05-23 21:14:48 -04:00
|
|
|
spark.stop()
|
2013-12-25 00:08:05 -05:00
|
|
|
if failure_count:
|
2018-03-08 06:38:34 -05:00
|
|
|
sys.exit(-1)
|
2013-12-25 00:08:05 -05:00
|
|
|
|
|
|
|
if __name__ == "__main__":
|
|
|
|
_test()
|