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
|
|
|
import numpy as np
|
|
|
|
from numpy import array
|
[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
|
|
|
|
2015-05-06 01:57:13 -04:00
|
|
|
from pyspark import RDD
|
2015-03-20 14:44:21 -04:00
|
|
|
from pyspark.mllib.common import callMLlibFunc, _py2java, _java2py, inherit_doc
|
2015-05-06 01:57:13 -04:00
|
|
|
from pyspark.mllib.linalg import SparseVector, Vectors, _convert_to_vector
|
2015-03-20 14:44:21 -04:00
|
|
|
from pyspark.mllib.util import Saveable, Loader
|
[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
|
|
|
|
2015-03-12 19:46:29 -04:00
|
|
|
__all__ = ['LabeledPoint', 'LinearModel',
|
|
|
|
'LinearRegressionModel', 'LinearRegressionWithSGD',
|
|
|
|
'RidgeRegressionModel', 'RidgeRegressionWithSGD',
|
2015-05-06 01:57:13 -04:00
|
|
|
'LassoModel', 'LassoWithSGD', 'IsotonicRegressionModel',
|
|
|
|
'IsotonicRegression']
|
2014-09-03 14:49:45 -04:00
|
|
|
|
|
|
|
|
[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
|
|
|
class LabeledPoint(object):
|
2014-08-06 15:58:24 -04:00
|
|
|
|
[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
|
|
|
"""
|
|
|
|
The features and labels of a data point.
|
|
|
|
|
2014-10-07 21:09:27 -04:00
|
|
|
:param label: Label for this data point.
|
2015-02-25 19:13:17 -05:00
|
|
|
:param features: Vector of features for this point (NumPy array,
|
|
|
|
list, pyspark.mllib.linalg.SparseVector, or scipy.sparse
|
|
|
|
column matrix)
|
|
|
|
|
|
|
|
Note: 'label' and 'features' are accessible as class attributes.
|
[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
|
|
|
"""
|
2014-08-06 15:58:24 -04:00
|
|
|
|
[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
|
|
|
def __init__(self, label, features):
|
2014-11-11 01:26:16 -05:00
|
|
|
self.label = float(label)
|
2014-09-19 18:01:11 -04:00
|
|
|
self.features = _convert_to_vector(features)
|
|
|
|
|
|
|
|
def __reduce__(self):
|
|
|
|
return (LabeledPoint, (self.label, self.features))
|
[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
|
|
|
|
2014-06-04 15:56:56 -04:00
|
|
|
def __str__(self):
|
2014-09-19 18:01:11 -04:00
|
|
|
return "(" + ",".join((str(self.label), str(self.features))) + ")"
|
|
|
|
|
|
|
|
def __repr__(self):
|
2014-11-11 01:26:16 -05:00
|
|
|
return "LabeledPoint(%s, %s)" % (self.label, self.features)
|
2014-06-04 15:56:56 -04:00
|
|
|
|
2013-12-25 00:08:05 -05:00
|
|
|
|
|
|
|
class LinearModel(object):
|
2014-08-06 15:58:24 -04:00
|
|
|
|
[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
|
|
|
"""A linear model that has a vector of coefficients and an intercept."""
|
2014-08-06 15:58:24 -04:00
|
|
|
|
[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
|
|
|
def __init__(self, weights, intercept):
|
2014-09-19 18:01:11 -04:00
|
|
|
self._coeff = _convert_to_vector(weights)
|
2014-11-11 01:26:16 -05:00
|
|
|
self._intercept = float(intercept)
|
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
|
|
|
@property
|
|
|
|
def weights(self):
|
|
|
|
return self._coeff
|
|
|
|
|
|
|
|
@property
|
|
|
|
def intercept(self):
|
|
|
|
return self._intercept
|
|
|
|
|
2014-10-06 17:05:45 -04:00
|
|
|
def __repr__(self):
|
2014-11-11 01:26:16 -05:00
|
|
|
return "(weights=%s, intercept=%r)" % (self._coeff, self._intercept)
|
2014-10-06 17:05:45 -04:00
|
|
|
|
[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
|
|
|
|
2015-02-20 05:31:32 -05:00
|
|
|
@inherit_doc
|
2013-12-25 00:08:05 -05:00
|
|
|
class LinearRegressionModelBase(LinearModel):
|
2014-08-06 15:58:24 -04:00
|
|
|
|
2013-12-25 00:08:05 -05:00
|
|
|
"""A linear regression model.
|
|
|
|
|
2014-09-19 18:01:11 -04:00
|
|
|
>>> lrmb = LinearRegressionModelBase(np.array([1.0, 2.0]), 0.1)
|
|
|
|
>>> abs(lrmb.predict(np.array([-1.03, 7.777])) - 14.624) < 1e-6
|
2013-12-25 00:08:05 -05:00
|
|
|
True
|
[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
|
|
|
>>> abs(lrmb.predict(SparseVector(2, {0: -1.03, 1: 7.777})) - 14.624) < 1e-6
|
|
|
|
True
|
2013-12-25 00:08:05 -05:00
|
|
|
"""
|
2014-08-06 15:58:24 -04:00
|
|
|
|
2013-12-25 00:08:05 -05:00
|
|
|
def predict(self, x):
|
2014-09-19 18:01:11 -04:00
|
|
|
"""
|
|
|
|
Predict the value of the dependent variable given a vector x
|
|
|
|
containing values for the independent variables.
|
|
|
|
"""
|
2014-11-11 01:26:16 -05:00
|
|
|
x = _convert_to_vector(x)
|
2014-09-19 18:01:11 -04:00
|
|
|
return self.weights.dot(x) + self.intercept
|
[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
|
|
|
|
2015-02-20 05:31:32 -05:00
|
|
|
@inherit_doc
|
2013-12-25 00:08:05 -05:00
|
|
|
class LinearRegressionModel(LinearRegressionModelBase):
|
2014-08-06 15:58:24 -04:00
|
|
|
|
2013-12-25 00:08:05 -05:00
|
|
|
"""A linear regression model derived from a least-squares fit.
|
|
|
|
|
[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
|
|
|
>>> from pyspark.mllib.regression import LabeledPoint
|
|
|
|
>>> data = [
|
|
|
|
... LabeledPoint(0.0, [0.0]),
|
|
|
|
... LabeledPoint(1.0, [1.0]),
|
|
|
|
... LabeledPoint(3.0, [2.0]),
|
|
|
|
... LabeledPoint(2.0, [3.0])
|
|
|
|
... ]
|
2015-04-21 20:49:55 -04:00
|
|
|
>>> lrm = LinearRegressionWithSGD.train(sc.parallelize(data), iterations=10,
|
|
|
|
... initialWeights=np.array([1.0]))
|
2014-09-19 18:01:11 -04:00
|
|
|
>>> abs(lrm.predict(np.array([0.0])) - 0) < 0.5
|
[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
|
|
|
True
|
2014-09-19 18:01:11 -04:00
|
|
|
>>> abs(lrm.predict(np.array([1.0])) - 1) < 0.5
|
[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
|
|
|
True
|
|
|
|
>>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5
|
|
|
|
True
|
2015-03-20 14:44:21 -04:00
|
|
|
>>> import os, tempfile
|
|
|
|
>>> path = tempfile.mkdtemp()
|
|
|
|
>>> lrm.save(sc, path)
|
|
|
|
>>> sameModel = LinearRegressionModel.load(sc, path)
|
|
|
|
>>> abs(sameModel.predict(np.array([0.0])) - 0) < 0.5
|
|
|
|
True
|
|
|
|
>>> abs(sameModel.predict(np.array([1.0])) - 1) < 0.5
|
|
|
|
True
|
|
|
|
>>> abs(sameModel.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5
|
|
|
|
True
|
|
|
|
>>> try:
|
|
|
|
... os.removedirs(path)
|
|
|
|
... except:
|
|
|
|
... pass
|
[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, SparseVector(1, {0: 0.0})),
|
|
|
|
... LabeledPoint(1.0, SparseVector(1, {0: 1.0})),
|
|
|
|
... LabeledPoint(3.0, SparseVector(1, {0: 2.0})),
|
|
|
|
... LabeledPoint(2.0, SparseVector(1, {0: 3.0}))
|
|
|
|
... ]
|
2015-04-21 20:49:55 -04:00
|
|
|
>>> lrm = LinearRegressionWithSGD.train(sc.parallelize(data), iterations=10,
|
|
|
|
... initialWeights=array([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
|
|
|
>>> abs(lrm.predict(array([0.0])) - 0) < 0.5
|
|
|
|
True
|
|
|
|
>>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5
|
|
|
|
True
|
2015-04-21 20:49:55 -04:00
|
|
|
>>> lrm = LinearRegressionWithSGD.train(sc.parallelize(data), iterations=10, step=1.0,
|
2015-03-25 16:38:33 -04:00
|
|
|
... miniBatchFraction=1.0, initialWeights=array([1.0]), regParam=0.1, regType="l2",
|
|
|
|
... intercept=True, validateData=True)
|
|
|
|
>>> abs(lrm.predict(array([0.0])) - 0) < 0.5
|
|
|
|
True
|
|
|
|
>>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5
|
|
|
|
True
|
2013-12-25 00:08:05 -05:00
|
|
|
"""
|
2015-03-20 14:44:21 -04:00
|
|
|
def save(self, sc, path):
|
|
|
|
java_model = sc._jvm.org.apache.spark.mllib.regression.LinearRegressionModel(
|
|
|
|
_py2java(sc, self._coeff), self.intercept)
|
|
|
|
java_model.save(sc._jsc.sc(), path)
|
|
|
|
|
|
|
|
@classmethod
|
|
|
|
def load(cls, sc, path):
|
|
|
|
java_model = sc._jvm.org.apache.spark.mllib.regression.LinearRegressionModel.load(
|
|
|
|
sc._jsc.sc(), path)
|
|
|
|
weights = _java2py(sc, java_model.weights())
|
|
|
|
intercept = java_model.intercept()
|
|
|
|
model = LinearRegressionModel(weights, intercept)
|
|
|
|
return 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
|
|
|
|
2014-09-19 18:01:11 -04:00
|
|
|
# train_func should take two parameters, namely data and initial_weights, and
|
|
|
|
# return the result of a call to the appropriate JVM stub.
|
|
|
|
# _regression_train_wrapper is responsible for setup and error checking.
|
2014-10-31 01:25:18 -04:00
|
|
|
def _regression_train_wrapper(train_func, modelClass, data, initial_weights):
|
2015-03-31 14:32:14 -04:00
|
|
|
from pyspark.mllib.classification import LogisticRegressionModel
|
2014-11-11 01:26:16 -05:00
|
|
|
first = data.first()
|
|
|
|
if not isinstance(first, LabeledPoint):
|
2015-04-20 13:44:09 -04:00
|
|
|
raise TypeError("data should be an RDD of LabeledPoint, but got %s" % type(first))
|
2015-03-20 17:18:18 -04:00
|
|
|
if initial_weights is None:
|
|
|
|
initial_weights = [0.0] * len(data.first().features)
|
2015-03-31 14:32:14 -04:00
|
|
|
if (modelClass == LogisticRegressionModel):
|
|
|
|
weights, intercept, numFeatures, numClasses = train_func(
|
|
|
|
data, _convert_to_vector(initial_weights))
|
|
|
|
return modelClass(weights, intercept, numFeatures, numClasses)
|
|
|
|
else:
|
|
|
|
weights, intercept = train_func(data, _convert_to_vector(initial_weights))
|
|
|
|
return modelClass(weights, intercept)
|
2014-09-19 18:01:11 -04:00
|
|
|
|
|
|
|
|
2013-12-25 00:08:05 -05:00
|
|
|
class LinearRegressionWithSGD(object):
|
2014-08-06 15:58:24 -04:00
|
|
|
|
2013-12-25 00:08:05 -05:00
|
|
|
@classmethod
|
2014-08-02 00:00:31 -04:00
|
|
|
def train(cls, data, iterations=100, step=1.0, miniBatchFraction=1.0,
|
2015-03-25 16:38:33 -04:00
|
|
|
initialWeights=None, regParam=0.0, regType=None, intercept=False,
|
|
|
|
validateData=True):
|
2014-08-02 00:00:31 -04:00
|
|
|
"""
|
|
|
|
Train a linear regression model on the given data.
|
|
|
|
|
2014-10-07 21:09:27 -04:00
|
|
|
:param data: The training data.
|
|
|
|
:param iterations: The number of iterations (default: 100).
|
|
|
|
:param step: The step parameter used in SGD
|
2014-08-02 00:00:31 -04:00
|
|
|
(default: 1.0).
|
2014-10-07 21:09:27 -04:00
|
|
|
:param miniBatchFraction: Fraction of data to be used for each SGD
|
2014-08-02 00:00:31 -04:00
|
|
|
iteration.
|
2014-10-07 21:09:27 -04:00
|
|
|
:param initialWeights: The initial weights (default: None).
|
2014-11-13 16:54:16 -05:00
|
|
|
:param regParam: The regularizer parameter (default: 0.0).
|
2014-10-07 21:09:27 -04:00
|
|
|
:param regType: The type of regularizer used for training
|
2014-08-02 00:00:31 -04:00
|
|
|
our model.
|
2014-10-06 17:08:40 -04:00
|
|
|
|
|
|
|
:Allowed values:
|
2014-11-13 16:54:16 -05:00
|
|
|
- "l1" for using L1 regularization (lasso),
|
|
|
|
- "l2" for using L2 regularization (ridge),
|
|
|
|
- None for no regularization
|
2014-10-06 17:08:40 -04:00
|
|
|
|
2014-11-13 16:54:16 -05:00
|
|
|
(default: None)
|
2014-10-06 17:08:40 -04:00
|
|
|
|
2015-03-25 16:38:33 -04:00
|
|
|
:param intercept: Boolean parameter which indicates the use
|
2014-08-02 00:00:31 -04:00
|
|
|
or not of the augmented representation for
|
|
|
|
training data (i.e. whether bias features
|
2015-02-20 05:31:32 -05:00
|
|
|
are activated or not). (default: False)
|
2015-03-25 16:38:33 -04:00
|
|
|
:param validateData: Boolean parameter which indicates if the
|
|
|
|
algorithm should validate data before training.
|
|
|
|
(default: True)
|
2014-08-02 00:00:31 -04:00
|
|
|
"""
|
2014-10-31 01:25:18 -04:00
|
|
|
def train(rdd, i):
|
2014-11-13 16:54:16 -05:00
|
|
|
return callMLlibFunc("trainLinearRegressionModelWithSGD", rdd, int(iterations),
|
|
|
|
float(step), float(miniBatchFraction), i, float(regParam),
|
2015-03-25 16:38:33 -04:00
|
|
|
regType, bool(intercept), bool(validateData))
|
2014-09-19 18:01:11 -04:00
|
|
|
|
2014-11-13 16:54:16 -05:00
|
|
|
return _regression_train_wrapper(train, LinearRegressionModel, data, initialWeights)
|
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
|
|
|
|
2015-02-20 05:31:32 -05:00
|
|
|
@inherit_doc
|
2013-12-25 00:08:05 -05:00
|
|
|
class LassoModel(LinearRegressionModelBase):
|
2014-08-06 15:58:24 -04:00
|
|
|
|
2013-12-25 00:08:05 -05:00
|
|
|
"""A linear regression model derived from a least-squares fit with an
|
|
|
|
l_1 penalty term.
|
|
|
|
|
[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
|
|
|
>>> from pyspark.mllib.regression import LabeledPoint
|
|
|
|
>>> data = [
|
|
|
|
... LabeledPoint(0.0, [0.0]),
|
|
|
|
... LabeledPoint(1.0, [1.0]),
|
|
|
|
... LabeledPoint(3.0, [2.0]),
|
|
|
|
... LabeledPoint(2.0, [3.0])
|
|
|
|
... ]
|
2015-04-21 20:49:55 -04:00
|
|
|
>>> lrm = LassoWithSGD.train(sc.parallelize(data), iterations=10, initialWeights=array([1.0]))
|
2014-09-19 18:01:11 -04:00
|
|
|
>>> abs(lrm.predict(np.array([0.0])) - 0) < 0.5
|
[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
|
|
|
True
|
2014-09-19 18:01:11 -04:00
|
|
|
>>> abs(lrm.predict(np.array([1.0])) - 1) < 0.5
|
[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
|
|
|
True
|
|
|
|
>>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5
|
|
|
|
True
|
2015-03-20 14:44:21 -04:00
|
|
|
>>> import os, tempfile
|
|
|
|
>>> path = tempfile.mkdtemp()
|
|
|
|
>>> lrm.save(sc, path)
|
|
|
|
>>> sameModel = LassoModel.load(sc, path)
|
|
|
|
>>> abs(sameModel.predict(np.array([0.0])) - 0) < 0.5
|
|
|
|
True
|
|
|
|
>>> abs(sameModel.predict(np.array([1.0])) - 1) < 0.5
|
|
|
|
True
|
|
|
|
>>> abs(sameModel.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5
|
|
|
|
True
|
|
|
|
>>> try:
|
|
|
|
... os.removedirs(path)
|
|
|
|
... except:
|
|
|
|
... pass
|
[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, SparseVector(1, {0: 0.0})),
|
|
|
|
... LabeledPoint(1.0, SparseVector(1, {0: 1.0})),
|
|
|
|
... LabeledPoint(3.0, SparseVector(1, {0: 2.0})),
|
|
|
|
... LabeledPoint(2.0, SparseVector(1, {0: 3.0}))
|
|
|
|
... ]
|
2015-04-21 20:49:55 -04:00
|
|
|
>>> lrm = LinearRegressionWithSGD.train(sc.parallelize(data), iterations=10,
|
|
|
|
... initialWeights=array([1.0]))
|
2014-09-19 18:01:11 -04:00
|
|
|
>>> abs(lrm.predict(np.array([0.0])) - 0) < 0.5
|
[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
|
|
|
True
|
|
|
|
>>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5
|
|
|
|
True
|
2015-04-21 20:49:55 -04:00
|
|
|
>>> lrm = LassoWithSGD.train(sc.parallelize(data), iterations=10, step=1.0,
|
2015-03-25 16:38:33 -04:00
|
|
|
... regParam=0.01, miniBatchFraction=1.0, initialWeights=array([1.0]), intercept=True,
|
|
|
|
... validateData=True)
|
|
|
|
>>> abs(lrm.predict(np.array([0.0])) - 0) < 0.5
|
|
|
|
True
|
|
|
|
>>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5
|
|
|
|
True
|
2013-12-25 00:08:05 -05:00
|
|
|
"""
|
2015-03-20 14:44:21 -04:00
|
|
|
def save(self, sc, path):
|
|
|
|
java_model = sc._jvm.org.apache.spark.mllib.regression.LassoModel(
|
|
|
|
_py2java(sc, self._coeff), self.intercept)
|
|
|
|
java_model.save(sc._jsc.sc(), path)
|
|
|
|
|
|
|
|
@classmethod
|
|
|
|
def load(cls, sc, path):
|
|
|
|
java_model = sc._jvm.org.apache.spark.mllib.regression.LassoModel.load(
|
|
|
|
sc._jsc.sc(), path)
|
|
|
|
weights = _java2py(sc, java_model.weights())
|
|
|
|
intercept = java_model.intercept()
|
|
|
|
model = LassoModel(weights, intercept)
|
|
|
|
return model
|
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
|
|
|
|
2013-12-25 00:08:05 -05:00
|
|
|
class LassoWithSGD(object):
|
2014-08-06 15:58:24 -04:00
|
|
|
|
2013-12-25 00:08:05 -05:00
|
|
|
@classmethod
|
2014-11-13 16:54:16 -05:00
|
|
|
def train(cls, data, iterations=100, step=1.0, regParam=0.01,
|
2015-03-25 16:38:33 -04:00
|
|
|
miniBatchFraction=1.0, initialWeights=None, intercept=False,
|
|
|
|
validateData=True):
|
2013-12-25 00:08:05 -05:00
|
|
|
"""Train a Lasso regression model on the given data."""
|
2014-10-31 01:25:18 -04:00
|
|
|
def train(rdd, i):
|
2014-11-13 16:54:16 -05:00
|
|
|
return callMLlibFunc("trainLassoModelWithSGD", rdd, int(iterations), float(step),
|
2015-03-25 16:38:33 -04:00
|
|
|
float(regParam), float(miniBatchFraction), i, bool(intercept),
|
|
|
|
bool(validateData))
|
2014-11-13 16:54:16 -05:00
|
|
|
|
2014-10-31 01:25:18 -04:00
|
|
|
return _regression_train_wrapper(train, LassoModel, data, initialWeights)
|
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
|
|
|
|
2015-02-20 05:31:32 -05:00
|
|
|
@inherit_doc
|
2013-12-25 00:08:05 -05:00
|
|
|
class RidgeRegressionModel(LinearRegressionModelBase):
|
2014-08-06 15:58:24 -04:00
|
|
|
|
2013-12-25 00:08:05 -05:00
|
|
|
"""A linear regression model derived from a least-squares fit with an
|
|
|
|
l_2 penalty term.
|
|
|
|
|
[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
|
|
|
>>> from pyspark.mllib.regression import LabeledPoint
|
|
|
|
>>> data = [
|
|
|
|
... LabeledPoint(0.0, [0.0]),
|
|
|
|
... LabeledPoint(1.0, [1.0]),
|
|
|
|
... LabeledPoint(3.0, [2.0]),
|
|
|
|
... LabeledPoint(2.0, [3.0])
|
|
|
|
... ]
|
2015-04-21 20:49:55 -04:00
|
|
|
>>> lrm = RidgeRegressionWithSGD.train(sc.parallelize(data), iterations=10,
|
|
|
|
... initialWeights=array([1.0]))
|
2014-09-19 18:01:11 -04:00
|
|
|
>>> abs(lrm.predict(np.array([0.0])) - 0) < 0.5
|
[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
|
|
|
True
|
2014-09-19 18:01:11 -04:00
|
|
|
>>> abs(lrm.predict(np.array([1.0])) - 1) < 0.5
|
[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
|
|
|
True
|
|
|
|
>>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5
|
|
|
|
True
|
2015-03-20 14:44:21 -04:00
|
|
|
>>> import os, tempfile
|
|
|
|
>>> path = tempfile.mkdtemp()
|
|
|
|
>>> lrm.save(sc, path)
|
|
|
|
>>> sameModel = RidgeRegressionModel.load(sc, path)
|
|
|
|
>>> abs(sameModel.predict(np.array([0.0])) - 0) < 0.5
|
|
|
|
True
|
|
|
|
>>> abs(sameModel.predict(np.array([1.0])) - 1) < 0.5
|
|
|
|
True
|
|
|
|
>>> abs(sameModel.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5
|
|
|
|
True
|
|
|
|
>>> try:
|
|
|
|
... os.removedirs(path)
|
|
|
|
... except:
|
|
|
|
... pass
|
[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, SparseVector(1, {0: 0.0})),
|
|
|
|
... LabeledPoint(1.0, SparseVector(1, {0: 1.0})),
|
|
|
|
... LabeledPoint(3.0, SparseVector(1, {0: 2.0})),
|
|
|
|
... LabeledPoint(2.0, SparseVector(1, {0: 3.0}))
|
|
|
|
... ]
|
2015-04-21 20:49:55 -04:00
|
|
|
>>> lrm = LinearRegressionWithSGD.train(sc.parallelize(data), iterations=10,
|
|
|
|
... initialWeights=array([1.0]))
|
2014-09-19 18:01:11 -04:00
|
|
|
>>> abs(lrm.predict(np.array([0.0])) - 0) < 0.5
|
[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
|
|
|
True
|
|
|
|
>>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5
|
|
|
|
True
|
2015-04-21 20:49:55 -04:00
|
|
|
>>> lrm = RidgeRegressionWithSGD.train(sc.parallelize(data), iterations=10, step=1.0,
|
2015-03-25 16:38:33 -04:00
|
|
|
... regParam=0.01, miniBatchFraction=1.0, initialWeights=array([1.0]), intercept=True,
|
|
|
|
... validateData=True)
|
|
|
|
>>> abs(lrm.predict(np.array([0.0])) - 0) < 0.5
|
|
|
|
True
|
|
|
|
>>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5
|
|
|
|
True
|
2013-12-25 00:08:05 -05:00
|
|
|
"""
|
2015-03-20 14:44:21 -04:00
|
|
|
def save(self, sc, path):
|
|
|
|
java_model = sc._jvm.org.apache.spark.mllib.regression.RidgeRegressionModel(
|
|
|
|
_py2java(sc, self._coeff), self.intercept)
|
|
|
|
java_model.save(sc._jsc.sc(), path)
|
|
|
|
|
|
|
|
@classmethod
|
|
|
|
def load(cls, sc, path):
|
|
|
|
java_model = sc._jvm.org.apache.spark.mllib.regression.RidgeRegressionModel.load(
|
|
|
|
sc._jsc.sc(), path)
|
|
|
|
weights = _java2py(sc, java_model.weights())
|
|
|
|
intercept = java_model.intercept()
|
|
|
|
model = RidgeRegressionModel(weights, intercept)
|
|
|
|
return 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
|
|
|
|
2013-12-25 00:08:05 -05:00
|
|
|
class RidgeRegressionWithSGD(object):
|
2014-08-06 15:58:24 -04:00
|
|
|
|
2013-12-25 00:08:05 -05:00
|
|
|
@classmethod
|
2014-11-13 16:54:16 -05:00
|
|
|
def train(cls, data, iterations=100, step=1.0, regParam=0.01,
|
2015-03-25 16:38:33 -04:00
|
|
|
miniBatchFraction=1.0, initialWeights=None, intercept=False,
|
|
|
|
validateData=True):
|
2013-12-25 00:08:05 -05:00
|
|
|
"""Train a ridge regression model on the given data."""
|
2014-10-31 01:25:18 -04:00
|
|
|
def train(rdd, i):
|
2014-11-13 16:54:16 -05:00
|
|
|
return callMLlibFunc("trainRidgeModelWithSGD", rdd, int(iterations), float(step),
|
2015-03-25 16:38:33 -04:00
|
|
|
float(regParam), float(miniBatchFraction), i, bool(intercept),
|
|
|
|
bool(validateData))
|
2014-09-19 18:01:11 -04:00
|
|
|
|
2014-11-13 16:54:16 -05:00
|
|
|
return _regression_train_wrapper(train, RidgeRegressionModel, data, initialWeights)
|
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
|
|
|
|
2015-05-06 01:57:13 -04:00
|
|
|
class IsotonicRegressionModel(Saveable, Loader):
|
|
|
|
|
|
|
|
"""Regression model for isotonic regression.
|
|
|
|
|
|
|
|
>>> data = [(1, 0, 1), (2, 1, 1), (3, 2, 1), (1, 3, 1), (6, 4, 1), (17, 5, 1), (16, 6, 1)]
|
|
|
|
>>> irm = IsotonicRegression.train(sc.parallelize(data))
|
|
|
|
>>> irm.predict(3)
|
|
|
|
2.0
|
|
|
|
>>> irm.predict(5)
|
|
|
|
16.5
|
|
|
|
>>> irm.predict(sc.parallelize([3, 5])).collect()
|
|
|
|
[2.0, 16.5]
|
|
|
|
>>> import os, tempfile
|
|
|
|
>>> path = tempfile.mkdtemp()
|
|
|
|
>>> irm.save(sc, path)
|
|
|
|
>>> sameModel = IsotonicRegressionModel.load(sc, path)
|
|
|
|
>>> sameModel.predict(3)
|
|
|
|
2.0
|
|
|
|
>>> sameModel.predict(5)
|
|
|
|
16.5
|
|
|
|
>>> try:
|
|
|
|
... os.removedirs(path)
|
|
|
|
... except OSError:
|
|
|
|
... pass
|
|
|
|
"""
|
|
|
|
|
|
|
|
def __init__(self, boundaries, predictions, isotonic):
|
|
|
|
self.boundaries = boundaries
|
|
|
|
self.predictions = predictions
|
|
|
|
self.isotonic = isotonic
|
|
|
|
|
|
|
|
def predict(self, x):
|
|
|
|
if isinstance(x, RDD):
|
|
|
|
return x.map(lambda v: self.predict(v))
|
|
|
|
return np.interp(x, self.boundaries, self.predictions)
|
|
|
|
|
|
|
|
def save(self, sc, path):
|
|
|
|
java_boundaries = _py2java(sc, self.boundaries.tolist())
|
|
|
|
java_predictions = _py2java(sc, self.predictions.tolist())
|
|
|
|
java_model = sc._jvm.org.apache.spark.mllib.regression.IsotonicRegressionModel(
|
|
|
|
java_boundaries, java_predictions, self.isotonic)
|
|
|
|
java_model.save(sc._jsc.sc(), path)
|
|
|
|
|
|
|
|
@classmethod
|
|
|
|
def load(cls, sc, path):
|
|
|
|
java_model = sc._jvm.org.apache.spark.mllib.regression.IsotonicRegressionModel.load(
|
|
|
|
sc._jsc.sc(), path)
|
|
|
|
py_boundaries = _java2py(sc, java_model.boundaryVector()).toArray()
|
|
|
|
py_predictions = _java2py(sc, java_model.predictionVector()).toArray()
|
|
|
|
return IsotonicRegressionModel(py_boundaries, py_predictions, java_model.isotonic)
|
|
|
|
|
|
|
|
|
|
|
|
class IsotonicRegression(object):
|
|
|
|
"""
|
|
|
|
Run IsotonicRegression algorithm to obtain isotonic regression model.
|
|
|
|
|
|
|
|
:param data: RDD of (label, feature, weight) tuples.
|
|
|
|
:param isotonic: Whether this is isotonic or antitonic.
|
|
|
|
"""
|
|
|
|
@classmethod
|
|
|
|
def train(cls, data, isotonic=True):
|
|
|
|
"""Train a isotonic regression model on the given data."""
|
|
|
|
boundaries, predictions = callMLlibFunc("trainIsotonicRegressionModel",
|
|
|
|
data.map(_convert_to_vector), bool(isotonic))
|
|
|
|
return IsotonicRegressionModel(boundaries.toArray(), predictions.toArray(), isotonic)
|
|
|
|
|
|
|
|
|
2013-12-25 00:08:05 -05:00
|
|
|
def _test():
|
|
|
|
import doctest
|
2014-10-31 01:25:18 -04:00
|
|
|
from pyspark import SparkContext
|
2014-11-11 01:26:16 -05:00
|
|
|
import pyspark.mllib.regression
|
|
|
|
globs = pyspark.mllib.regression.__dict__.copy()
|
2015-04-21 20:49:55 -04:00
|
|
|
globs['sc'] = SparkContext('local[2]', 'PythonTest', batchSize=2)
|
2014-05-25 20:15:01 -04:00
|
|
|
(failure_count, test_count) = doctest.testmod(globs=globs, optionflags=doctest.ELLIPSIS)
|
2013-12-25 00:08:05 -05:00
|
|
|
globs['sc'].stop()
|
|
|
|
if failure_count:
|
|
|
|
exit(-1)
|
|
|
|
|
|
|
|
if __name__ == "__main__":
|
|
|
|
_test()
|