In-place updates, reduce number of transposes, and vectorize operations in OnlineLDA implementation.
Author: Feynman Liang <fliang@databricks.com>
Closes#7454 from feynmanliang/OnlineLDA-perf-improvements and squashes the following commits:
78b0f5a [Feynman Liang] Make in-place variables vals, fix BLAS error
7f62a55 [Feynman Liang] --amend
c62cb1e [Feynman Liang] Outer product for stats, revert Range slicing
aead650 [Feynman Liang] Range slice, in-place update, reduce transposes
Add support for saving and loading LDA both the local and distributed versions.
Author: MechCoder <manojkumarsivaraj334@gmail.com>
Closes#6948 from MechCoder/lda_save_load and squashes the following commits:
49bcdce [MechCoder] minor style fixes
cc14054 [MechCoder] minor
4587d1d [MechCoder] Minor changes
c753122 [MechCoder] Load and save the model in private methods
2782326 [MechCoder] [SPARK-5989] Model save/load for LDA
Created since tags for methods in mllib.classification
Author: petz2000 <petz2000@gmail.com>
Closes#7371 from petz2000/add_since_mllib.classification and squashes the following commits:
39fe291 [petz2000] Removed whitespace in block comment
c9b1e03 [petz2000] Removed @since tags again from protected and private methods
cd759b6 [petz2000] Added @since tags to methods
Author: Holden Karau <holden@pigscanfly.ca>
Closes#7553 from holdenk/SPARK-9204-add-default-params-test-to-linear-regression and squashes the following commits:
630ba19 [Holden Karau] style fix
faa08a3 [Holden Karau] Add default params test for linearyregression suite
This exposes the SparkR:::glm() and SparkR:::predict() APIs. It was necessary to change RFormula to silently drop the label column if it was missing from the input dataset, which is kind of a hack but necessary to integrate with the Pipeline API.
The umbrella design doc for MLlib + SparkR integration can be viewed here: https://docs.google.com/document/d/10NZNSEurN2EdWM31uFYsgayIPfCFHiuIu3pCWrUmP_c/edit
mengxr
Author: Eric Liang <ekl@databricks.com>
Closes#7483 from ericl/spark-8774 and squashes the following commits:
3dfac0c [Eric Liang] update
17ef516 [Eric Liang] more comments
1753a0f [Eric Liang] make glm generic
b0f50f8 [Eric Liang] equivalence test
550d56d [Eric Liang] export methods
c015697 [Eric Liang] second pass
117949a [Eric Liang] comments
5afbc67 [Eric Liang] test label columns
6b7f15f [Eric Liang] Fri Jul 17 14:20:22 PDT 2015
3a63ae5 [Eric Liang] Fri Jul 17 13:41:52 PDT 2015
ce61367 [Eric Liang] Fri Jul 17 13:41:17 PDT 2015
0299c59 [Eric Liang] Fri Jul 17 13:40:32 PDT 2015
e37603f [Eric Liang] Fri Jul 17 12:15:03 PDT 2015
d417d0c [Eric Liang] Merge remote-tracking branch 'upstream/master' into spark-8774
29a2ce7 [Eric Liang] Merge branch 'spark-8774-1' into spark-8774
d1959d2 [Eric Liang] clarify comment
2db68aa [Eric Liang] second round of comments
dc3c943 [Eric Liang] address comments
5765ec6 [Eric Liang] fix style checks
1f361b0 [Eric Liang] doc
d33211b [Eric Liang] r support
fb0826b [Eric Liang] [SPARK-8774] Add R model formula with basic support as a transformer
Fix BLAS.gemm to update matrix C when alpha==0 and beta!=1
Also include unit tests to verify the fix.
mengxr brkyvz
Author: Meihua Wu <meihuawu@umich.edu>
Closes#7503 from rotationsymmetry/fix_BLAS_gemm and squashes the following commits:
fce199c [Meihua Wu] Fix BLAS.gemm to update C when alpha==0 and beta!=1
Python API for the KS-test
Statistics.kolmogorovSmirnovTest(data, distName, *params)
I'm not quite sure how to support the callable function since it is not serializable.
Author: MechCoder <manojkumarsivaraj334@gmail.com>
Closes#7430 from MechCoder/spark-8996 and squashes the following commits:
2dd009d [MechCoder] minor
021d233 [MechCoder] Remove one wrapper and other minor stuff
49d07ab [MechCoder] [SPARK-8996] [MLlib] Python API for Kolmogorov-Smirnov Test
Modifying Vector, DenseVector, and SparseVector to implement argmax functionality. This work is to set the stage for changes to be done in Spark-7423.
Author: George Dittmar <georgedittmar@gmail.com>
Author: George <dittmar@Georges-MacBook-Pro.local>
Author: dittmarg <george.dittmar@webtrends.com>
Author: Xiangrui Meng <meng@databricks.com>
Closes#6112 from GeorgeDittmar/SPARK-7422 and squashes the following commits:
3e0a939 [George Dittmar] Merge pull request #1 from mengxr/SPARK-7422
127dec5 [Xiangrui Meng] update argmax impl
2ea6a55 [George Dittmar] Added MimaExcludes for Vectors.argmax
98058f4 [George Dittmar] Merge branch 'master' of github.com:apache/spark into SPARK-7422
5fd9380 [George Dittmar] fixing style check error
42341fb [George Dittmar] refactoring arg max check to better handle zero values
b22af46 [George Dittmar] Fixing spaces between commas in unit test
f2eba2f [George Dittmar] Cleaning up unit tests to be fewer lines
aa330e3 [George Dittmar] Fixing some last if else spacing issues
ac53c55 [George Dittmar] changing dense vector argmax unit test to be one line call vs 2
d5b5423 [George Dittmar] Fixing code style and updating if logic on when to check for zero values
ee1a85a [George Dittmar] Cleaning up unit tests a bit and modifying a few cases
3ee8711 [George Dittmar] Fixing corner case issue with zeros in the active values of the sparse vector. Updated unit tests
b1f059f [George Dittmar] Added comment before we start arg max calculation. Updated unit tests to cover corner cases
f21dcce [George Dittmar] commit
af17981 [dittmarg] Initial work fixing bug that was made clear in pr
eeda560 [George] Fixing SparseVector argmax function to ignore zero values while doing the calculation.
4526acc [George] Merge branch 'master' of github.com:apache/spark into SPARK-7422
df9538a [George] Added argmax to sparse vector and added unit test
3cffed4 [George] Adding unit tests for argmax functions for Dense and Sparse vectors
04677af [George] initial work on adding argmax to Vector and SparseVector
I Implemented the KMeans API for spark.ml Pipelines. But it doesn't include clustering abstractions for spark.ml (SPARK-7610). It would fit for another issues. And I'll try it later, since we are trying to add the hierarchical clustering algorithms in another issue. Thanks.
[SPARK-7879] KMeans API for spark.ml Pipelines - ASF JIRA https://issues.apache.org/jira/browse/SPARK-7879
Author: Yu ISHIKAWA <yuu.ishikawa@gmail.com>
Closes#6756 from yu-iskw/SPARK-7879 and squashes the following commits:
be752de [Yu ISHIKAWA] Add assertions
a14939b [Yu ISHIKAWA] Fix the dashed line's length in pyspark.ml.rst
4c61693 [Yu ISHIKAWA] Remove the test about whether "features" and "prediction" columns exist or not in Python
fb2417c [Yu ISHIKAWA] Use getInt, instead of get
f397be4 [Yu ISHIKAWA] Switch the comparisons.
ca78b7d [Yu ISHIKAWA] Add the Scala docs about the constraints of each parameter.
effc650 [Yu ISHIKAWA] Using expertSetParam and expertGetParam
c8dc6e6 [Yu ISHIKAWA] Remove an unnecessary test
19a9d63 [Yu ISHIKAWA] Include spark.ml.clustering to python tests
1abb19c [Yu ISHIKAWA] Add the statements about spark.ml.clustering into pyspark.ml.rst
f8338bc [Yu ISHIKAWA] Add the placeholders in Python
4a03003 [Yu ISHIKAWA] Test for contains in Python
6566c8b [Yu ISHIKAWA] Use `get`, instead of `apply`
288e8d5 [Yu ISHIKAWA] Using `contains` to check the column names
5a7d574 [Yu ISHIKAWA] Renamce `validateInitializationMode` to `validateInitMode` and remove throwing exception
97cfae3 [Yu ISHIKAWA] Fix the type of return value of `KMeans.copy`
e933723 [Yu ISHIKAWA] Remove the default value of seed from the Model class
978ee2c [Yu ISHIKAWA] Modify the docs of KMeans, according to mllib's KMeans
2ec80bc [Yu ISHIKAWA] Fit on 1 line
e186be1 [Yu ISHIKAWA] Make a few variables, setters and getters be expert ones
b2c205c [Yu ISHIKAWA] Rename the method `getInitializationSteps` to `getInitSteps` and `setInitializationSteps` to `setInitSteps` in Scala and Python
f43f5b4 [Yu ISHIKAWA] Rename the method `getInitializationMode` to `getInitMode` and `setInitializationMode` to `setInitMode` in Scala and Python
3cb5ba4 [Yu ISHIKAWA] Modify the description about epsilon and the validation
4fa409b [Yu ISHIKAWA] Add a comment about the default value of epsilon
2f392e1 [Yu ISHIKAWA] Make some variables `final` and Use `IntParam` and `DoubleParam`
19326f8 [Yu ISHIKAWA] Use `udf`, instead of callUDF
4d2ad1e [Yu ISHIKAWA] Modify the indentations
0ae422f [Yu ISHIKAWA] Add a test for `setParams`
4ff7913 [Yu ISHIKAWA] Add "ml.clustering" to `javacOptions` in SparkBuild.scala
11ffdf1 [Yu ISHIKAWA] Use `===` and the variable
220a176 [Yu ISHIKAWA] Set a random seed in the unit testing
92c3efc [Yu ISHIKAWA] Make the points for a test be fewer
c758692 [Yu ISHIKAWA] Modify the parameters of KMeans in Python
6aca147 [Yu ISHIKAWA] Add some unit testings to validate the setter methods
687cacc [Yu ISHIKAWA] Alias mllib.KMeans as MLlibKMeans in KMeansSuite.scala
a4dfbef [Yu ISHIKAWA] Modify the last brace and indentations
5bedc51 [Yu ISHIKAWA] Remve an extra new line
444c289 [Yu ISHIKAWA] Add the validation for `runs`
e41989c [Yu ISHIKAWA] Modify how to validate `initStep`
7ea133a [Yu ISHIKAWA] Change how to validate `initMode`
7991e15 [Yu ISHIKAWA] Add a validation for `k`
c2df35d [Yu ISHIKAWA] Make `predict` private
93aa2ff [Yu ISHIKAWA] Use `withColumn` in `transform`
d3a79f7 [Yu ISHIKAWA] Remove the inhefited docs
e9532e1 [Yu ISHIKAWA] make `parentModel` of KMeansModel private
8559772 [Yu ISHIKAWA] Remove the `paramMap` parameter of KMeans
6684850 [Yu ISHIKAWA] Rename `initializationSteps` to `initSteps`
99b1b96 [Yu ISHIKAWA] Rename `initializationMode` to `initMode`
79ea82b [Yu ISHIKAWA] Modify the parameters of KMeans docs
6569bcd [Yu ISHIKAWA] Change how to set the default values with `setDefault`
20a795a [Yu ISHIKAWA] Change how to set the default values with `setDefault`
11c2a12 [Yu ISHIKAWA] Limit the imports
badb481 [Yu ISHIKAWA] Alias spark.mllib.{KMeans, KMeansModel}
f80319a [Yu ISHIKAWA] Rebase mater branch and add copy methods
85d92b1 [Yu ISHIKAWA] Add `KMeans.setPredictionCol`
aa9469d [Yu ISHIKAWA] Fix a python test suite error caused by python 3.x
c2d6bcb [Yu ISHIKAWA] ADD Java test suites of the KMeans API for spark.ml Pipeline
598ed2e [Yu ISHIKAWA] Implement the KMeans API for spark.ml Pipelines in Python
63ad785 [Yu ISHIKAWA] Implement the KMeans API for spark.ml Pipelines in Scala
Broadcast of ensemble models in transformImpl before call to predict
Author: Bryan Cutler <bjcutler@us.ibm.com>
Closes#6300 from BryanCutler/bcast-ensemble-models-7127 and squashes the following commits:
86e73de [Bryan Cutler] [SPARK-7127] Replaced deprecated callUDF with udf
40a139d [Bryan Cutler] Merge branch 'master' into bcast-ensemble-models-7127
9afad56 [Bryan Cutler] [SPARK-7127] Simplified calls by overriding transformImpl and using broadcasted model in callUDF to make prediction
1f34be4 [Bryan Cutler] [SPARK-7127] Removed accidental newline
171a6ce [Bryan Cutler] [SPARK-7127] Used modelAccessor parameter in predictImpl to access broadcasted model
6fd153c [Bryan Cutler] [SPARK-7127] Applied broadcasting to remaining ensemble models
aaad77b [Bryan Cutler] [SPARK-7127] Removed abstract class for broadcasting model, instead passing a prediction function as param to transform
83904bb [Bryan Cutler] [SPARK-7127] Adding broadcast of model before prediction in RandomForestClassifier
Make the definition of residuals in Spark consistent with literature. We have been using `prediction - label` for residuals, but literature usually defines `residual = label - prediction`.
Author: Feynman Liang <fliang@databricks.com>
Closes#7435 from feynmanliang/SPARK-9090-Fix-LinearRegressionSummary-Residuals and squashes the following commits:
f4b39d8 [Feynman Liang] Fix doc
bc12a92 [Feynman Liang] Tweak EnsembleTestHelper and SquaredError residuals
63f0d60 [Feynman Liang] Fix definition of residual
jira: https://issues.apache.org/jira/browse/SPARK-9062
Currently output type of Tokenizer is Array(String, false), which is not compatible with Word2Vec and Other transformers since their input type is Array(String, true). Seq[String] in udf will be treated as Array(String, true) by default.
I'm not sure what's the recommended way for Tokenizer to handle the null value in the input. Any suggestion will be welcome.
Author: Yuhao Yang <hhbyyh@gmail.com>
Closes#7414 from hhbyyh/tokenizer and squashes the following commits:
c01bd7a [Yuhao Yang] change output type of tokenizer
fix wrong annotation of RFormula.formula
Author: Yanbo Liang <ybliang8@gmail.com>
Closes#7470 from yanboliang/RFormula and squashes the following commits:
61f1919 [Yanbo Liang] fix wrong annotation
Measure lower and upper bounds for task time and use them for validation. This PR also implements `Stopwatch.toString`. This suite should finish in less than 1 second.
jkbradley pwendell
Author: Xiangrui Meng <meng@databricks.com>
Closes#7457 from mengxr/SPARK-9126 and squashes the following commits:
4b40faa [Xiangrui Meng] simplify tests
739f5bd [Xiangrui Meng] do not assert on time taken by Thread.sleep()
This PR copies the RandomForest implementation from spark.mllib to spark.ml. Note that this includes the DecisionTree implementation, but not the GradientBoostedTrees one (which will come later).
I essentially copied a minimal amount of code to spark.ml, removed the use of bins (and only used splits), and modified code only as much as necessary to get it to compile. The spark.ml implementation still uses some spark.mllib classes (privately), which can be moved in future PRs.
This refactoring will be helpful in extending the node representation to include more information, such as class probabilities.
Specifically:
* Copied code from spark.mllib to spark.ml:
* mllib.tree.DecisionTree, mllib.tree.RandomForest copied to ml.tree.impl.RandomForest (main implementation)
* NodeIdCache (needed to use splits instead of bins)
* TreePoint (use splits instead of bins)
* Added ml.tree.LearningNode used in RandomForest training (needed vars)
* Removed bins from implementation, and only used splits
* Small fix in JavaDecisionTreeRegressorSuite
CC: mengxr manishamde codedeft chouqin
Author: Joseph K. Bradley <joseph@databricks.com>
Closes#7294 from jkbradley/dt-move-impl and squashes the following commits:
48749be [Joseph K. Bradley] cleanups based on code review, mostly style
bea9703 [Joseph K. Bradley] scala style fixes. added some scala doc
4e6d2a4 [Joseph K. Bradley] removed unnecessary use of copyValues, setParent for trees
9a4d721 [Joseph K. Bradley] cleanups. removed InfoGainStats from ml, using old one for now.
836e7d4 [Joseph K. Bradley] Fixed test suite failures
bd5e063 [Joseph K. Bradley] fixed bucketizing issue
0df3759 [Joseph K. Bradley] Need to remove use of Bucketizer
d5224a9 [Joseph K. Bradley] modified tree and forest to use moved impl
cc01823 [Joseph K. Bradley] still editing RF to get it to work
19143fb [Joseph K. Bradley] More progress, but not done yet. Rebased with master after 1.4 release.
Add stopwatches for easy instrumentation of MLlib algorithms. This is based on the `TimeTracker` used in decision trees. The distributed version uses Spark accumulator. jkbradley
Author: Xiangrui Meng <meng@databricks.com>
Closes#7415 from mengxr/SPARK-9018 and squashes the following commits:
40b4347 [Xiangrui Meng] == -> ===
c477745 [Xiangrui Meng] address Joseph's comments
f981a49 [Xiangrui Meng] add stopwatches
This implements minimal R formula support as a feature transformer. Both numeric and string labels are supported, but features must be numeric for now.
cc mengxr
Author: Eric Liang <ekl@databricks.com>
Closes#7381 from ericl/spark-8774-1 and squashes the following commits:
d1959d2 [Eric Liang] clarify comment
2db68aa [Eric Liang] second round of comments
dc3c943 [Eric Liang] address comments
5765ec6 [Eric Liang] fix style checks
1f361b0 [Eric Liang] doc
fb0826b [Eric Liang] [SPARK-8774] Add R model formula with basic support as a transformer
Fixes implementation of `explainedVariance` and `r2` to be consistent with their definitions as described in [SPARK-9005](https://issues.apache.org/jira/browse/SPARK-9005).
Author: Feynman Liang <fliang@databricks.com>
Closes#7361 from feynmanliang/SPARK-9005-RegressionMetrics-bugs and squashes the following commits:
f1112fc [Feynman Liang] Add explainedVariance formula
1a3d098 [Feynman Liang] SROwen code review comments
08a0e1b [Feynman Liang] Fix pyspark tests
db8605a [Feynman Liang] Style fix
bde9761 [Feynman Liang] Fix RegressionMetrics tests, relax assumption predictor is unbiased
c235de0 [Feynman Liang] Fix RegressionMetrics tests
4c4e56f [Feynman Liang] Fix RegressionMetrics computation of explainedVariance and r2
This allows Kmeans to be initialized using an existing set of cluster centers provided as a KMeansModel object. This mode of initialization performs a single run.
Author: FlytxtRnD <meethu.mathew@flytxt.com>
Closes#6737 from FlytxtRnD/Kmeans-8018 and squashes the following commits:
94b56df [FlytxtRnD] style correction
ef95ee2 [FlytxtRnD] style correction
c446c58 [FlytxtRnD] documentation and numRuns warning change
06d13ef [FlytxtRnD] numRuns corrected
d12336e [FlytxtRnD] numRuns variable modifications
07f8554 [FlytxtRnD] remove setRuns from setIntialModel
e721dfe [FlytxtRnD] Merge remote-tracking branch 'upstream/master' into Kmeans-8018
242ead1 [FlytxtRnD] corrected == to === in assert
714acb5 [FlytxtRnD] added numRuns
60c8ce2 [FlytxtRnD] ignore runs parameter and initialModel test suite changed
582e6d9 [FlytxtRnD] Merge remote-tracking branch 'upstream/master' into Kmeans-8018
3f5fc8e [FlytxtRnD] test case modified and one runs condition added
cd5dc5c [FlytxtRnD] Merge remote-tracking branch 'upstream/master' into Kmeans-8018
16f1b53 [FlytxtRnD] Merge branch 'Kmeans-8018', remote-tracking branch 'upstream/master' into Kmeans-8018
e9c35d7 [FlytxtRnD] Remove getInitialModel and match cluster count criteria
6959861 [FlytxtRnD] Accept initial cluster centers in KMeans
I implemented the Python API for LDA. But I didn't implemented a method for `LDAModel.describeTopics()`, beause it's a little hard to implement it now. And adding document about that and an example code would fit for another issue.
TODO: LDAModel.describeTopics() in Python must be also implemented. But it would be nice to fit for another issue. Implementing it is a little hard, since the return value of `describeTopics` in Scala consists of Tuple classes.
Author: Yu ISHIKAWA <yuu.ishikawa@gmail.com>
Closes#6791 from yu-iskw/SPARK-6259 and squashes the following commits:
6855f59 [Yu ISHIKAWA] LDA inherits object
28bd165 [Yu ISHIKAWA] Change the place of testing code
d7a332a [Yu ISHIKAWA] Remove the doc comment about the optimizer's default value
083e226 [Yu ISHIKAWA] Add the comment about the supported values and the default value of `optimizer`
9f8bed8 [Yu ISHIKAWA] Simplify casting
faa9764 [Yu ISHIKAWA] Add some comments for the LDA paramters
98f645a [Yu ISHIKAWA] Remove the interface for `describeTopics`. Because it is not implemented.
57ac03d [Yu ISHIKAWA] Remove the unnecessary import in Python unit testing
73412c3 [Yu ISHIKAWA] Fix the typo
2278829 [Yu ISHIKAWA] Fix the indentation
39514ec [Yu ISHIKAWA] Modify how to cast the input data
8117e18 [Yu ISHIKAWA] Fix the validation problems by `lint-scala`
77fd1b7 [Yu ISHIKAWA] Not use LabeledPoint
68f0653 [Yu ISHIKAWA] Support some parameters for `ALS.train()` in Python
25ef2ac [Yu ISHIKAWA] Resolve conflicts with rebasing
Add predictProbabilities to Naive Bayes, return class probabilities.
Continues https://github.com/apache/spark/pull/6761
Author: Sean Owen <sowen@cloudera.com>
Closes#7376 from srowen/SPARK-4362 and squashes the following commits:
23d5a76 [Sean Owen] Fix model.labels -> model.theta
95d91fb [Sean Owen] Check that predicted probabilities sum to 1
b32d1c8 [Sean Owen] Add predictProbabilities to Naive Bayes, return class probabilities
Removed private[ml] from Generated documentation
Author: Vinod K C <vinod.kc@huawei.com>
Closes#7367 from vinodkc/fix_sharedparmascodegen and squashes the following commits:
4fa3c8f [Vinod K C] Adding auto generated code
7e19025 [Vinod K C] Removed private[ml]
Made default impl of Params.validateParams empty
CC mengxr
Author: Joseph K. Bradley <joseph@databricks.com>
Closes#7349 from jkbradley/pipeline-small-cleanups and squashes the following commits:
4e0f013 [Joseph K. Bradley] small cleanups after SPARK-5956
Add parallel PrefixSpan algorithm and test file.
Support non-temporal sequences.
Author: zhangjiajin <zhangjiajin@huawei.com>
Author: zhang jiajin <zhangjiajin@huawei.com>
Closes#7258 from zhangjiajin/master and squashes the following commits:
ca9c4c8 [zhangjiajin] Modified the code according to the review comments.
574e56c [zhangjiajin] Add new object LocalPrefixSpan, and do some optimization.
ba5df34 [zhangjiajin] Fix a Scala style error.
4c60fb3 [zhangjiajin] Fix some Scala style errors.
1dd33ad [zhangjiajin] Modified the code according to the review comments.
89bc368 [zhangjiajin] Fixed a Scala style error.
a2eb14c [zhang jiajin] Delete PrefixspanSuite.scala
951fd42 [zhang jiajin] Delete Prefixspan.scala
575995f [zhangjiajin] Modified the code according to the review comments.
91fd7e6 [zhangjiajin] Add new algorithm PrefixSpan and test file.
This contribution is my original work and I license it to the project under it's open source license.
Author: jose.cambronero <jose.cambronero@cloudera.com>
Closes#6994 from josepablocam/master and squashes the following commits:
bbb30b1 [jose.cambronero] renamed KSTestResult to KolmogorovSmirnovTestResult, to stay consistent with method name
0d0c201 [jose.cambronero] kstTest -> kolmogorovSmirnovTest in statistics.md
1f56371 [jose.cambronero] changed ksTest in public API to kolmogorovSmirnovTest for clarity
a48ae7b [jose.cambronero] refactor code to account for serializable RealDistribution. Reuse testOneSample( _, cdf)
1bb44bd [jose.cambronero] style and doc changes. Factored out ks test into 2 separate tests
2ec2aa6 [jose.cambronero] initialize to stdnormal when no params passed (and log). Change unit tests to approximate equivalence rather than strict
a4bc0c7 [jose.cambronero] changed ksTest(data, distName) to ksTest(data, distName, params*) after api discussions. Changed tests and docs accordingly
7e66f57 [jose.cambronero] copied implementation note to public api docs, and added @see for links to wiki info
e760ebd [jose.cambronero] line length changes to fit style check
3288e42 [jose.cambronero] addressed style changes, correctness change to simpler approach, and fixed edge case for foldLeft in searchOneSampleCandidates when a partition is empty
9026895 [jose.cambronero] addressed style changes, correctness change to simpler approach, and fixed edge case for foldLeft in searchOneSampleCandidates when a partition is empty
1226b30 [jose.cambronero] reindent multi-line lambdas, prior intepretation of style guide was wrong on my part
9c0f1af [jose.cambronero] additional style changes incorporated and added documentation to mllib statistics docs
3f81ad2 [jose.cambronero] renamed ks1 sample test for clarity
992293b [jose.cambronero] Style changes as per comments and added implementation note explaining the distributed approach.
6a4784f [jose.cambronero] specified what distributions are available for the convenience method ksTest(data, name) (solely standard normal)
4b8ba61 [jose.cambronero] fixed off by 1/N in cases when post-constant adjustment ecdf is above cdf, but prior to adj it was below
0b5e8ec [jose.cambronero] changed KS one sample test to perform just 1 distributed pass (in addition to the sorting pass), operates on each partition separately. Implementation of Sandy Ryza's algorithm
16b5c4c [jose.cambronero] renamed dat to data and eliminated recalc of RDD size by sharing as argument between empirical and evalOneSampleP
c18dc66 [jose.cambronero] removed ksTestOpt from API and changed comments in HypothesisTestSuite accordingly
f6951b6 [jose.cambronero] changed style and some comments based on feedback from pull request
b9cff3a [jose.cambronero] made small changes to pass style check
ce8e9a1 [jose.cambronero] added kstest testing in HypothesisTestSuite
4da189b [jose.cambronero] added user facing ks test functions
c659ea1 [jose.cambronero] created KS test class
13dfe4d [jose.cambronero] created test result class for ks test
Author: rahulpalamuttam <rahulpalamut@gmail.com>
Closes#7341 from rahulpalamuttam/TaggingMLlibfpm and squashes the following commits:
bef2843 [rahulpalamuttam] fix @since tags in mmlib.fpm
cd86252 [rahulpalamuttam] Add @since tags to mllib.fpm
Author: Jonathan Alter <jonalter@users.noreply.github.com>
Closes#7093 from jonalter/SPARK-7977 and squashes the following commits:
ccd44cc [Jonathan Alter] Changed println to log in ThreadingSuite
7fcac3e [Jonathan Alter] Reverting to println in ThreadingSuite
10724b6 [Jonathan Alter] Changing some printlns to logs in tests
eeec1e7 [Jonathan Alter] Merge branch 'master' of github.com:apache/spark into SPARK-7977
0b1dcb4 [Jonathan Alter] More println cleanup
aedaf80 [Jonathan Alter] Merge branch 'master' of github.com:apache/spark into SPARK-7977
925fd98 [Jonathan Alter] Merge branch 'master' of github.com:apache/spark into SPARK-7977
0c16fa3 [Jonathan Alter] Replacing some printlns with logs
45c7e05 [Jonathan Alter] Merge branch 'master' of github.com:apache/spark into SPARK-7977
5c8e283 [Jonathan Alter] Allowing println in audit-release examples
5b50da1 [Jonathan Alter] Allowing printlns in example files
ca4b477 [Jonathan Alter] Merge branch 'master' of github.com:apache/spark into SPARK-7977
83ab635 [Jonathan Alter] Fixing new printlns
54b131f [Jonathan Alter] Merge branch 'master' of github.com:apache/spark into SPARK-7977
1cd8a81 [Jonathan Alter] Removing some unnecessary comments and printlns
b837c3a [Jonathan Alter] Disallowing println
Cleanup tests from SPARK 8700.
Author: Holden Karau <holden@pigscanfly.ca>
Closes#7335 from holdenk/SPARK-8913-cleanup-tests-from-SPARK-8700-logistic-regression-r2-really-logistic-regression-this-time and squashes the following commits:
e5e2c5f [Holden Karau] Simplify LogisticRegression suite to use Vector <-> Vector comparisions instead of comparing element by element
Adds results (e.g. objective value at each iteration, residuals) on training and user-specified test sets for LinearRegressionModel.
Notes to Reviewers:
* Are the `*TrainingResults` and `Results` classes too specialized for `LinearRegressionModel`? Where would be an appropriate level of abstraction?
* Please check `transient` annotations are correct; the datasets should not be copied and kept during serialization.
* Any thoughts on `RDD`s versus `DataFrame`s? If using `DataFrame`s, suggested schemas for each intermediate step? Also, how to create a "local DataFrame" without a `sqlContext`?
Author: Feynman Liang <fliang@databricks.com>
Closes#7099 from feynmanliang/SPARK-8538 and squashes the following commits:
d219fa4 [Feynman Liang] Update docs
4a42680 [Feynman Liang] Change Summary to hold values, move transient annotations down to metrics and predictions DF
6300031 [Feynman Liang] Code review changes
0a5e762 [Feynman Liang] Fix build error
e71102d [Feynman Liang] Merge branch 'master' into SPARK-8538
3367489 [Feynman Liang] Merge branch 'master' into SPARK-8538
70f267c [Feynman Liang] Make TrainingSummary transient and remove Serializable from *Summary and RegressionMetrics
1d9ea42 [Feynman Liang] Fix failing Java test
a65dfda [Feynman Liang] Make TrainingSummary and metrics serializable, prediction dataframe transient
0a605d8 [Feynman Liang] Replace Params from LinearRegression*Summary with private constructor vals
c2fe835 [Feynman Liang] Optimize imports
02d8a70 [Feynman Liang] Add Params to LinearModel*Summary, refactor tests and add test for evaluate()
8f999f4 [Feynman Liang] Refactor from jkbradley code review
072e948 [Feynman Liang] Style
509ae36 [Feynman Liang] Use DFs and localize serialization to LinearRegressionModel
9509c79 [Feynman Liang] Fix imports
b2bbaa3 [Feynman Liang] Refactored LinearRegressionResults API to be more private
ffceaec [Feynman Liang] Merge branch 'master' into SPARK-8538
1cedb2b [Feynman Liang] Add test for decreasing objective trace
dab0aff [Feynman Liang] Add LinearRegressionTrainingResults tests, make test suite code copy+pasteable
97b0a81 [Feynman Liang] Add LinearRegressionModel.evaluate() to get results on test sets
dc51bce [Feynman Liang] Style guide fixes
521f397 [Feynman Liang] Use RDD[(Double, Double)] instead of DF
2ff5710 [Feynman Liang] Add training results and model summary to ML LinearRegression
Simplify model weight assertions to use vector comparision, switch to using absTol when comparing with 0.0 intercepts
Author: Holden Karau <holden@pigscanfly.ca>
Closes#7327 from holdenk/SPARK-8913-cleanup-tests-from-SPARK-8700-logistic-regression and squashes the following commits:
5bac185 [Holden Karau] Simplify model weight assertions to use vector comparision, switch to using absTol when comparing with 0.0 intercepts
jira: https://issues.apache.org/jira/browse/SPARK-8703
Converts a text document to a sparse vector of token counts.
I can further add an estimator to extract vocabulary from corpus if that's appropriate.
Author: Yuhao Yang <hhbyyh@gmail.com>
Closes#7084 from hhbyyh/countVectorization and squashes the following commits:
5f3f655 [Yuhao Yang] text change
24728e4 [Yuhao Yang] style improvement
576728a [Yuhao Yang] rename to model and some fix
1deca28 [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into countVectorization
99b0c14 [Yuhao Yang] undo extension from HashingTF
12c2dc8 [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into countVectorization
7ee1c31 [Yuhao Yang] extends HashingTF
809fb59 [Yuhao Yang] minor fix for ut
7c61fb3 [Yuhao Yang] add countVectorizer
This PR fixes the converter for Python DataFrame, especially for DecimalType
Closes#7106
Author: Davies Liu <davies@databricks.com>
Closes#7131 from davies/decimal_python and squashes the following commits:
4d3c234 [Davies Liu] Merge branch 'master' of github.com:apache/spark into decimal_python
20531d6 [Davies Liu] Merge branch 'master' of github.com:apache/spark into decimal_python
7d73168 [Davies Liu] fix conflit
6cdd86a [Davies Liu] Merge branch 'master' of github.com:apache/spark into decimal_python
7104e97 [Davies Liu] improve type infer
9cd5a21 [Davies Liu] run python tests with SPARK_PREPEND_CLASSES
829a05b [Davies Liu] fix UDT in python
c99e8c5 [Davies Liu] fix mima
c46814a [Davies Liu] convert decimal for Python DataFrames
Distribute expensive portions of computation for Gaussian mixture components (in particular, pre-computation of `MultivariateGaussian.rootSigmaInv`, the inverse covariance matrix and covariance determinant) across executors. Repost of PR#4654.
Notes for reviewers:
* What should be the policy for when to distribute computation. Always? When numClusters > threshold? User-specified param?
TODO:
* Performance testing and comparison for large number of clusters
Author: Feynman Liang <fliang@databricks.com>
Closes#7166 from feynmanliang/GMM_parallel_mixtures and squashes the following commits:
4f351fa [Feynman Liang] Update heuristic and scaladoc
5ea947e [Feynman Liang] Fix parallelization logic
00eb7db [Feynman Liang] Add helper method for GMM's M step, remove distributeGaussians flag
e7c8127 [Feynman Liang] Add distributeGaussians flag and tests
1da3c7f [Feynman Liang] Distribute mixtures
Adds FPGrowth.generateAssociationRules to public API for generating association rules after mining frequent itemsets.
Author: Feynman Liang <fliang@databricks.com>
Closes#7271 from feynmanliang/SPARK-8877 and squashes the following commits:
83b8baf [Feynman Liang] Add API Doc
867abff [Feynman Liang] Add FPGrowth.generateAssociationRules and change access modifiers for AssociationRules
All compressed sensing applications, and some of the regression use-cases will have better result by turning the feature scaling off. However, if we implement this naively by training the dataset without doing any standardization, the rate of convergency will not be good. This can be implemented by still standardizing the training dataset but we penalize each component differently to get effectively the same objective function but a better numerical problem. As a result, for those columns with high variances, they will be penalized less, and vice versa. Without this, since all the features are standardized, so they will be penalized the same.
In R, there is an option for this.
`standardize`
Logical flag for x variable standardization, prior to fitting the model sequence. The coefficients are always returned on the original scale. Default is standardize=TRUE. If variables are in the same units already, you might not wish to standardize. See details below for y standardization with family="gaussian".
+cc holdenk mengxr jkbradley
Author: DB Tsai <dbt@netflix.com>
Closes#7080 from dbtsai/lors and squashes the following commits:
877e6c7 [DB Tsai] repahse the doc
7cf45f2 [DB Tsai] address feedback
78d75c9 [DB Tsai] small change
c2c9e60 [DB Tsai] style
6e1a8e0 [DB Tsai] first commit
Author: Kashif Rasul <kashif.rasul@gmail.com>
Closes#7269 from kashif/SPARK-8872 and squashes the following commits:
2d5457f [Kashif Rasul] added R code for FP Int type
3de6808 [Kashif Rasul] added verification results from R for FPGrowthSuite
In LinearRegression and LogisticRegression, we use Breeze's optimizers (LBFGS and OWLQN). We check the State.value to see the current objective. However, Breeze's documentation makes it sound like value and adjustedValue differ for some optimizers, possibly including OWLQN: 26faf62286/math/src/main/scala/breeze/optimize/FirstOrderMinimizer.scala (L36)
If that is the case, then we should use adjustedValue instead of value. This is relevant to SPARK-8538 and SPARK-8539, where we will provide the objective trace to the user.
Author: DB Tsai <dbt@netflix.com>
Closes#7245 from dbtsai/SPARK-8845 and squashes the following commits:
fa4c91e [DB Tsai] address feedback
e6caac1 [DB Tsai] java style multiline comment
b10c574 [DB Tsai] address feedback
c9ff81e [DB Tsai] first commit
Add std, mean to StandardScalerModel
getVectors, findSynonyms to Word2Vec Model
setFeatures and getFeatures to hashingTF
Author: MechCoder <manojkumarsivaraj334@gmail.com>
Closes#7086 from MechCoder/missing_model_methods and squashes the following commits:
9fbae90 [MechCoder] Add type
6e3d6b2 [MechCoder] [SPARK-8704] Add missing methods in StandardScaler (ML and PySpark)
Distributed generation of single-consequent association rules from a RDD of frequent itemsets. Tests referenced against `R`'s implementation of A Priori in [arules](http://cran.r-project.org/web/packages/arules/index.html).
Author: Feynman Liang <fliang@databricks.com>
Closes#7005 from feynmanliang/fp-association-rules-distributed and squashes the following commits:
466ced0 [Feynman Liang] Refactor AR generation impl
73c1cff [Feynman Liang] Make rule attributes public, remove numTransactions from FreqItemset
80f63ff [Feynman Liang] Change default confidence and optimize imports
04cf5b5 [Feynman Liang] Code review with @mengxr, add R to tests
0cc1a6a [Feynman Liang] Java compatibility test
f3c14b5 [Feynman Liang] Fix MiMa test
764375e [Feynman Liang] Fix tests
1187307 [Feynman Liang] Almost working tests
b20779b [Feynman Liang] Working implementation
5395c4e [Feynman Liang] Fix imports
2d34405 [Feynman Liang] Partial implementation of distributed ar
83ace4b [Feynman Liang] Local rule generation without pruning complete
69c2c87 [Feynman Liang] Working local implementation, now to parallelize../..
4e1ec9a [Feynman Liang] Pull FreqItemsets out, refactor type param, tests
69ccedc [Feynman Liang] First implementation of association rule generation
Add numNodes and depth to treeModels, add treeWeights to ensemble Models.
Add __repr__ to all models.
Author: MechCoder <manojkumarsivaraj334@gmail.com>
Closes#7095 from MechCoder/missing_methods_tree and squashes the following commits:
23b08be [MechCoder] private [spark]
38a0860 [MechCoder] rename pyTreeWeights to javaTreeWeights
6d16ad8 [MechCoder] Fix Python 3 Error
47d7023 [MechCoder] Use np.allclose and treeEnsembleModel -> TreeEnsembleMethods
819098c [MechCoder] [SPARK-8711] [ML] Add additional methods ot PySpark ML tree models
Add Java unit test for PCA transformer
Author: Yanbo Liang <ybliang8@gmail.com>
Closes#7184 from yanboliang/spark-8788 and squashes the following commits:
9d1a2af [Yanbo Liang] address comments
b34451f [Yanbo Liang] Add Java unit test for PCA transformer
See the jira https://issues.apache.org/jira/browse/SPARK-5562
Author: Alok Singh <singhal@Aloks-MacBook-Pro.local>
Author: Alok Singh <singhal@aloks-mbp.usca.ibm.com>
Author: Alok Singh <“singhal@us.ibm.com”>
Closes#7064 from aloknsingh/aloknsingh_SPARK-5562 and squashes the following commits:
259a0a7 [Alok Singh] change as per the comments by @jkbradley
be48491 [Alok Singh] [SPARK-5562][MLlib] re-order import in alphabhetical order
c01311b [Alok Singh] [SPARK-5562][MLlib] fix the newline typo
b271c8a [Alok Singh] [SPARK-5562][Mllib] As per github discussion with jkbradley. We would like to simply things.
7c06251 [Alok Singh] [SPARK-5562][MLlib] modified the JavaLDASuite for test passing
c710cb6 [Alok Singh] fix the scala code style to have space after :
2572a08 [Alok Singh] [SPARK-5562][MLlib] change the import xyz._ to the import xyz.{c1, c2} ..
ab55fbf [Alok Singh] [SPARK-5562][MLlib] Change as per Sean Owen's comments https://github.com/apache/spark/pull/7064/files#diff-9236d23975e6f5a5608ffc81dfd79146
9f4f9ea [Alok Singh] [SPARK-5562][MLlib] LDA should handle empty document.
This reverts commit 25f574eb9a. After speaking to some users and developers, we realized that FP-growth doesn't meet the requirement for frequent sequence mining. PrefixSpan (SPARK-6487) would be the correct algorithm for it. feynmanliang
Author: Xiangrui Meng <meng@databricks.com>
Closes#7240 from mengxr/SPARK-7212.revert and squashes the following commits:
2b3d66b [Xiangrui Meng] Revert "[SPARK-7212] [MLLIB] Add sequence learning flag"
Author: Joshi <rekhajoshm@gmail.com>
Author: Rekha Joshi <rekhajoshm@gmail.com>
Closes#5992 from rekhajoshm/fix/SPARK-7137 and squashes the following commits:
8c42b57 [Joshi] update checkInputColumn to print more info if needed
33ddd2e [Joshi] update checkInputColumn to print more info if needed
acf3e17 [Joshi] update checkInputColumn to print more info if needed
8993c0e [Joshi] SPARK-7137: Add checkInputColumn back to Params and print more info
e3677c9 [Rekha Joshi] Merge pull request #1 from apache/master
Author: Yu ISHIKAWA <yuu.ishikawa@gmail.com>
Closes#6821 from yu-iskw/SPARK-7104 and squashes the following commits:
975136b [Yu ISHIKAWA] Organize import
0ef58b6 [Yu ISHIKAWA] Use rmtree, instead of removedirs
cb21653 [Yu ISHIKAWA] Add an explicit type for `Word2VecModelWrapper.save`
1d468ef [Yu ISHIKAWA] [SPARK-7104][MLlib] Support model save/load in Python's Word2Vec
GrandientDescent can receive convergence tolerance value. Default value is 0.0.
When loss value becomes less than the tolerance which is set by user, iteration is terminated.
Author: lewuathe <lewuathe@me.com>
Closes#3636 from Lewuathe/gd-convergence-tolerance and squashes the following commits:
0b8a9a8 [lewuathe] Update doc
ce91b15 [lewuathe] Merge branch 'master' into gd-convergence-tolerance
4f22c2b [lewuathe] Modify based on SPARK-1503
5e47b82 [lewuathe] Merge branch 'master' into gd-convergence-tolerance
abadb7e [lewuathe] Fix LassoSuite
8fadebd [lewuathe] Fix failed unit tests
ee5de46 [lewuathe] Merge branch 'master' into gd-convergence-tolerance
8313ba2 [lewuathe] Fix styles
0ead94c [lewuathe] Merge branch 'master' into gd-convergence-tolerance
a94cfd5 [lewuathe] Modify some styles
3aef0a2 [lewuathe] Modify converged logic to do relative comparison
f7b19d5 [lewuathe] [SPARK-3382] Clarify comparison logic
e6c9cd2 [lewuathe] [SPARK-3382] Compare with the diff of solution vector
4b125d2 [lewuathe] [SPARK3382] Fix scala style
e7c10dd [lewuathe] [SPARK-3382] format improvements
f867eea [lewuathe] [SPARK-3382] Modify warning message statements
b9d5e61 [lewuathe] [SPARK-3382] should compare diff inside loss history and convergence tolerance
5433f71 [lewuathe] [SPARK-3382] GradientDescent convergence tolerance
Matrices allow zeros to be stored in values. Sometimes a method is handy to check if the numNonZeros are same as number of Active values.
Author: MechCoder <manojkumarsivaraj334@gmail.com>
Closes#6904 from MechCoder/nnz_matrix and squashes the following commits:
252c6b7 [MechCoder] Add to MiMa excludes
e2390f5 [MechCoder] Use count instead of foreach
2f62b2f [MechCoder] Add to MiMa excludes
d6e96ef [MechCoder] [SPARK-8479] Add numNonzeros and numActives to linalg.Matrices
JIRA: https://issues.apache.org/jira/browse/SPARK-8708
Previously the partitions of ratings are only based on the given products. So if the `usersProducts` given for prediction contains only few products or even one product, the generated ratings will be pushed into few or single partition and can't use high parallelism.
The following codes are the example reported in the JIRA. Because it asks the predictions for users on product 2. There is only one partition in the result.
>>> r1 = (1, 1, 1.0)
>>> r2 = (1, 2, 2.0)
>>> r3 = (2, 1, 2.0)
>>> r4 = (2, 2, 2.0)
>>> r5 = (3, 1, 1.0)
>>> ratings = sc.parallelize([r1, r2, r3, r4, r5], 5)
>>> users = ratings.map(itemgetter(0)).distinct()
>>> model = ALS.trainImplicit(ratings, 1, seed=10)
>>> predictions_for_2 = model.predictAll(users.map(lambda u: (u, 2)))
>>> predictions_for_2.glom().map(len).collect()
[0, 0, 3, 0, 0]
This PR uses user and product instead of only product to partition the ratings.
Author: Liang-Chi Hsieh <viirya@gmail.com>
Author: Liang-Chi Hsieh <viirya@appier.com>
Closes#7121 from viirya/mfm_fix_partition and squashes the following commits:
779946d [Liang-Chi Hsieh] Calculate approximate numbers of users and products in one pass.
4336dc2 [Liang-Chi Hsieh] Merge remote-tracking branch 'upstream/master' into mfm_fix_partition
83e56c1 [Liang-Chi Hsieh] Instead of additional join, use the numbers of users and products to decide how to perform join.
b534dc8 [Liang-Chi Hsieh] Paritition ratings based on both users and products.
I added the code,
// see [SPARK-8647], this achieves the needed constant hash code without constant no.
override def hashCode(): Int = this.getClass.getName.hashCode()
does getting the constant hash code as per jira
Author: Alok Singh <singhal@Aloks-MacBook-Pro.local>
Closes#7146 from aloknsingh/aloknsingh_SPARK-8647 and squashes the following commits:
e58bccf [Alok Singh] [SPARK-8647][MLlib] to avoid the class derivation issues, change the constant hashCode to override def hashCode(): Int = classOf[MatrixUDT].getName.hashCode()
43cdb89 [Alok Singh] [SPARK-8647][MLlib] Potential issue with constant hashCode
I've updated default values in comments, documentation, and in the command line builder to be 1g based on comments in the JIRA. I've also updated most usages to point at a single variable defined in the Utils.scala and JavaUtils.java files. This wasn't possible in all cases (R, shell scripts etc.) but usage in most code is now pointing at the same place.
Please let me know if I've missed anything.
Will the spark-shell use the value within the command line builder during instantiation?
Author: Ilya Ganelin <ilya.ganelin@capitalone.com>
Closes#7132 from ilganeli/SPARK-3071 and squashes the following commits:
4074164 [Ilya Ganelin] String fix
271610b [Ilya Ganelin] Merge branch 'SPARK-3071' of github.com:ilganeli/spark into SPARK-3071
273b6e9 [Ilya Ganelin] Test fix
fd67721 [Ilya Ganelin] Update JavaUtils.java
26cc177 [Ilya Ganelin] test fix
e5db35d [Ilya Ganelin] Fixed test failure
39732a1 [Ilya Ganelin] merge fix
a6f7deb [Ilya Ganelin] Created default value for DRIVER MEM in Utils that's now used in almost all locations instead of setting manually in each
09ad698 [Ilya Ganelin] Update SubmitRestProtocolSuite.scala
19b6f25 [Ilya Ganelin] Missed one doc update
2698a3d [Ilya Ganelin] Updated default value for driver memory
'>' symbols removed from comments in LogisticRegressionSuite.scala, for ease of copypaste
also single-lined the multiline commands (is this desirable, or does it violate style?)
Author: Rosstin <asterazul@gmail.com>
Closes#7167 from Rosstin/SPARK-8660-2 and squashes the following commits:
f4b9bc8 [Rosstin] SPARK-8660 restored character limit on multiline comments in LogisticRegressionSuite.scala
fe6b112 [Rosstin] SPARK-8660 > symbols removed from LogisticRegressionSuite.scala for easy of copypaste
39ddd50 [Rosstin] Merge branch 'master' of github.com:apache/spark into SPARK-8661
5a05dee [Rosstin] SPARK-8661 for LinearRegressionSuite.scala, changed javadoc-style comments to regular multiline comments to make it easier to copy-paste the R code.
bb9a4b1 [Rosstin] Merge branch 'master' of github.com:apache/spark into SPARK-8660
242aedd [Rosstin] SPARK-8660, changed comment style from JavaDoc style to normal multiline comment in order to make copypaste into R easier, in file classification/LogisticRegressionSuite.scala
2cd2985 [Rosstin] Merge branch 'master' of github.com:apache/spark into SPARK-8639
21ac1e5 [Rosstin] Merge branch 'master' of github.com:apache/spark into SPARK-8639
6c18058 [Rosstin] fixed minor typos in docs/README.md and docs/api.md
Rename DiscreteCosineTransformer and related classes to DCT.
Author: Feynman Liang <fliang@databricks.com>
Closes#7138 from feynmanliang/dct-features and squashes the following commits:
e547b3e [Feynman Liang] Fix renaming bug
9d5c9e4 [Feynman Liang] Lowercase JavaDCTSuite variable
f9a8958 [Feynman Liang] Remove old files
f8fe794 [Feynman Liang] Merge branch 'master' into dct-features
894d0b2 [Feynman Liang] Rename DiscreteCosineTransformer to DCT
433dbc7 [Feynman Liang] Test refactoring
91e9636 [Feynman Liang] Style guide and test helper refactor
b5ac19c [Feynman Liang] Use Vector types, add Java test
530983a [Feynman Liang] Tests for other numeric datatypes
195d7aa [Feynman Liang] Implement support for arbitrary numeric types
95d4939 [Feynman Liang] Working DCT for 1D Doubles
I'm sorry that I made https://github.com/apache/spark/pull/6949 closed by mistake.
I pushed codes again.
And, I added a test code.
>
There is a bug that `U.numCols() = self.nCols` in `IndexedRowMatrix.computeSVD()`
It should have been `U.numCols() = k = svd.U.numCols()`
>
```
self = U * sigma * V.transpose
(m x n) = (m x n) * (k x k) * (k x n) //ASIS
-->
(m x n) = (m x k) * (k x k) * (k x n) //TOBE
```
Author: lee19 <lee19@live.co.kr>
Closes#6953 from lee19/MLlibBugfix and squashes the following commits:
c1812a0 [lee19] [SPARK-8563] [MLlib] Used nRows instead of numRows() to reduce a burden.
4b9803b [lee19] [SPARK-8563] [MLlib] Fixed a build error.
c2ccd89 [lee19] Added a unit test that validates matrix sizes of svd for [SPARK-8563][MLlib]
8373424 [lee19] [SPARK-8563][MLlib] Fixed a bug so that IndexedRowMatrix.computeSVD().U.numCols = k
Changed GBTRegressor so it does NOT threshold the prediction. Added test which fails with bug but works after fix.
CC: feynmanliang mengxr
Author: Joseph K. Bradley <joseph@databricks.com>
Closes#7134 from jkbradley/gbrt-fix and squashes the following commits:
613b90e [Joseph K. Bradley] Changed GBTRegressor so it does NOT threshold the prediction
jira: https://issues.apache.org/jira/browse/SPARK-7514
Add a popular scaling method to feature component, which is commonly known as min-max normalization or Rescaling.
Core function is,
Normalized(x) = (x - min) / (max - min) * scale + newBase
where `newBase` and `scale` are parameters (type Double) of the `VectorTransformer`. `newBase` is the new minimum number for the features, and `scale` controls the ranges after transformation. This is a little complicated than the basic MinMax normalization, yet it provides flexibility so that users can control the range more specifically. like [0.1, 0.9] in some NN application.
For case that `max == min`, 0.5 is used as the raw value. (0.5 * scale + newBase)
I'll add UT once the design got settled ( and this is not considered as too naive)
reference:
http://en.wikipedia.org/wiki/Feature_scalinghttp://stn.spotfire.com/spotfire_client_help/index.htm#norm/norm_scale_between_0_and_1.htm
Author: Yuhao Yang <hhbyyh@gmail.com>
Closes#6039 from hhbyyh/minMaxNorm and squashes the following commits:
f942e9f [Yuhao Yang] add todo for metadata
8b37bbc [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into minMaxNorm
4894dbc [Yuhao Yang] add copy
fa2989f [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into minMaxNorm
29db415 [Yuhao Yang] add clue and minor adjustment
5b8f7cc [Yuhao Yang] style fix
9b133d0 [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into minMaxNorm
22f20f2 [Yuhao Yang] style change and bug fix
747c9bb [Yuhao Yang] add ut and remove mllib version
a5ba0aa [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into minMaxNorm
585cc07 [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into minMaxNorm
1c6dcb1 [Yuhao Yang] minor change
0f1bc80 [Yuhao Yang] add MinMaxScaler to ml
8e7436e [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into minMaxNorm
3663165 [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into minMaxNorm
1247c27 [Yuhao Yang] some comments improvement
d285a19 [Yuhao Yang] initial checkin for minMaxNorm
Implementation and tests for Discrete Cosine Transformer.
Author: Feynman Liang <fliang@databricks.com>
Closes#6894 from feynmanliang/dct-features and squashes the following commits:
433dbc7 [Feynman Liang] Test refactoring
91e9636 [Feynman Liang] Style guide and test helper refactor
b5ac19c [Feynman Liang] Use Vector types, add Java test
530983a [Feynman Liang] Tests for other numeric datatypes
195d7aa [Feynman Liang] Implement support for arbitrary numeric types
95d4939 [Feynman Liang] Working DCT for 1D Doubles
Add PCA transformer for ML pipeline
Author: Yanbo Liang <ybliang8@gmail.com>
Closes#7065 from yanboliang/spark-8664 and squashes the following commits:
4afae45 [Yanbo Liang] address comments
e9effd7 [Yanbo Liang] Add PCA transformer
for mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala, changed javadoc-style comments to regular multiline comments, to make copy-pasting R code more simple
Author: Rosstin <asterazul@gmail.com>
Closes#7098 from Rosstin/SPARK-8661 and squashes the following commits:
5a05dee [Rosstin] SPARK-8661 for LinearRegressionSuite.scala, changed javadoc-style comments to regular multiline comments to make it easier to copy-paste the R code.
bb9a4b1 [Rosstin] Merge branch 'master' of github.com:apache/spark into SPARK-8660
242aedd [Rosstin] SPARK-8660, changed comment style from JavaDoc style to normal multiline comment in order to make copypaste into R easier, in file classification/LogisticRegressionSuite.scala
2cd2985 [Rosstin] Merge branch 'master' of github.com:apache/spark into SPARK-8639
21ac1e5 [Rosstin] Merge branch 'master' of github.com:apache/spark into SPARK-8639
6c18058 [Rosstin] fixed minor typos in docs/README.md and docs/api.md
Converted JavaDoc style comments in mllib/src/test/scala/org/apache/spark/ml/classification/LogisticRegressionSuite.scala to regular multiline comments, to make copy-pasting R commands easier.
Author: Rosstin <asterazul@gmail.com>
Closes#7096 from Rosstin/SPARK-8660 and squashes the following commits:
242aedd [Rosstin] SPARK-8660, changed comment style from JavaDoc style to normal multiline comment in order to make copypaste into R easier, in file classification/LogisticRegressionSuite.scala
2cd2985 [Rosstin] Merge branch 'master' of github.com:apache/spark into SPARK-8639
21ac1e5 [Rosstin] Merge branch 'master' of github.com:apache/spark into SPARK-8639
6c18058 [Rosstin] fixed minor typos in docs/README.md and docs/api.md
Follow up of [SPARK-8356](https://issues.apache.org/jira/browse/SPARK-8356) and #6902.
Removes the unit test for the now deprecated ```callUdf```
Unit test in SQLQuerySuite now uses ```udf``` instead of ```callUDF```
Replaced ```callUDF``` by ```udf``` where possible in mllib
Author: BenFradet <benjamin.fradet@gmail.com>
Closes#6993 from BenFradet/SPARK-8575 and squashes the following commits:
26f5a7a [BenFradet] 2 spaces instead of 1
1ddb452 [BenFradet] renamed initUDF in order to be consistent in OneVsRest
48ca15e [BenFradet] used vector type tag for udf call in VectorIndexer
0ebd0da [BenFradet] replace the now deprecated callUDF by udf in VectorIndexer
8013409 [BenFradet] replaced the now deprecated callUDF by udf in Predictor
94345b5 [BenFradet] unifomized udf calls in ProbabilisticClassifier
1305492 [BenFradet] uniformized udf calls in Classifier
a672228 [BenFradet] uniformized udf calls in OneVsRest
49e4904 [BenFradet] Revert "removal of the unit test for the now deprecated callUdf"
bbdeaf3 [BenFradet] fixed syntax for init udf in OneVsRest
fe2a10b [BenFradet] callUDF => udf in ProbabilisticClassifier
0ea30b3 [BenFradet] callUDF => udf in Classifier where possible
197ec82 [BenFradet] callUDF => udf in OneVsRest
84d6780 [BenFradet] modified unit test in SQLQuerySuite to use udf instead of callUDF
477709f [BenFradet] removal of the unit test for the now deprecated callUdf
Python support for Power Iteration Clustering
https://issues.apache.org/jira/browse/SPARK-5962
Author: Yanbo Liang <ybliang8@gmail.com>
Closes#6992 from yanboliang/pyspark-pic and squashes the following commits:
6b03d82 [Yanbo Liang] address comments
4be4423 [Yanbo Liang] Python support for Power Iteration Clustering
Support mining of ordered frequent item sequences.
Author: Feynman Liang <fliang@databricks.com>
Closes#6997 from feynmanliang/fp-sequence and squashes the following commits:
7c14e15 [Feynman Liang] Improve scalatests with R code and Seq
0d3e4b6 [Feynman Liang] Fix python test
ce987cb [Feynman Liang] Backwards compatibility aux constructor
34ef8f2 [Feynman Liang] Fix failing test due to reverse orderering
f04bd50 [Feynman Liang] Naming, add ordered to FreqItemsets, test ordering using Seq
648d4d4 [Feynman Liang] Test case for frequent item sequences
252a36a [Feynman Liang] Add sequence learning flag
Spark's tests currently depend on `mockito-all`, which bundles Hamcrest and Objenesis classes. Instead, it should depend on `mockito-core`, which declares those libraries as Maven dependencies. This is necessary in order to fix a dependency conflict that leads to a NoSuchMethodError when using certain Hamcrest matchers.
See https://github.com/mockito/mockito/wiki/Declaring-mockito-dependency for more details.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#7061 from JoshRosen/mockito-core-instead-of-all and squashes the following commits:
70eccbe [Josh Rosen] Depend on mockito-core instead of mockito-all.
Add a param to disable linear feature scaling (to be implemented later in linear & logistic regression). Done as a seperate PR so we can use same param & not conflict while working on the sub-tasks.
Author: Holden Karau <holden@pigscanfly.ca>
Closes#7024 from holdenk/SPARK-8522-Disable-Linear_featureScaling-Spark-8613-Add-param and squashes the following commits:
ce8931a [Holden Karau] Regenerate the sharedParams code
fa6427e [Holden Karau] update text for standardization param.
7b24a2b [Holden Karau] generate the new standardization param
3c190af [Holden Karau] Add the standardization param to sharedparamscodegen
Keep the same naming conventions for PythonMLLibAPI.
Only the following three functions is different from others
```scala
trainNaiveBayes
trainGaussianMixture
trainWord2Vec
```
So change them to
```scala
trainNaiveBayesModel
trainGaussianMixtureModel
trainWord2VecModel
```
It does not affect any users and public APIs, only to make better understand for developer and code hacker.
Author: Yanbo Liang <ybliang8@gmail.com>
Closes#7011 from yanboliang/py-mllib-api-rename and squashes the following commits:
771ffec [Yanbo Liang] rename some functions of PythonMLLibAPI
fix LabeledPoint parser when there is a whitespace between label and features vector, e.g.
(y, [x1, x2, x3])
Author: Oleksiy Dyagilev <oleksiy_dyagilev@epam.com>
Closes#6954 from fe2s/SPARK-8525 and squashes the following commits:
0755b9d [Oleksiy Dyagilev] [SPARK-8525][MLLIB] addressing comment, removing dep on commons-lang
c1abc2b [Oleksiy Dyagilev] [SPARK-8525][MLLIB] fix LabeledPoint parser when there is a whitespace on specific position
It is useful to generate linear data for easy testing of linear models and in general. Scala already has it. This is just a wrapper around the Scala code.
Author: MechCoder <manojkumarsivaraj334@gmail.com>
Closes#6715 from MechCoder/generate_linear_input and squashes the following commits:
6182884 [MechCoder] Minor changes
8bda047 [MechCoder] Minor style fixes
0f1053c [MechCoder] [SPARK-8265] Add LinearDataGenerator to pyspark.mllib.utils
Author: Holden Karau <holden@pigscanfly.ca>
Closes#6927 from holdenk/SPARK-7888-Be-able-to-disable-intercept-in-Linear-Regression-in-ML-package and squashes the following commits:
0ad384c [Holden Karau] Add MiMa excludes
4016fac [Holden Karau] Switch to wild card import, remove extra blank lines
ae5baa8 [Holden Karau] CR feedback, move the fitIntercept down rather than changing ymean and etc above
f34971c [Holden Karau] Fix some more long lines
319bd3f [Holden Karau] Fix long lines
3bb9ee1 [Holden Karau] Update the regression suite tests
7015b9f [Holden Karau] Our code performs the same with R, except we need more than one data point but that seems reasonable
0b0c8c0 [Holden Karau] fix the issue with the sample R code
e2140ba [Holden Karau] Add a test, it fails!
5e84a0b [Holden Karau] Write out thoughts and use the correct trait
91ffc0a [Holden Karau] more murh
006246c [Holden Karau] murp?
Author: Holden Karau <holden@pigscanfly.ca>
Closes#6331 from holdenk/SPARK-7781-GradientBoostedTrees.trainRegressor-missing-max-bins and squashes the following commits:
2894695 [Holden Karau] remove extra blank line
2573e8d [Holden Karau] Update the scala side of the pythonmllibapi and make the test a bit nicer too
3a09170 [Holden Karau] add maxBins to to the train method as well
af7f274 [Holden Karau] Add maxBins to GradientBoostedTrees.trainRegressor and correctly mention the default of 32 in other places where it mentioned 100
Implementation of n-gram feature transformer for ML.
Author: Feynman Liang <fliang@databricks.com>
Closes#6887 from feynmanliang/ngram-featurizer and squashes the following commits:
d2c839f [Feynman Liang] Make n > input length yield empty output
9fadd36 [Feynman Liang] Add empty and corner test cases, fix names and spaces
fe93873 [Feynman Liang] Implement n-gram feature transformer
Updated `Attribute.fromStructField` to allow any `NumericType`, rather than just `DoubleType`, and added unit tests for a few of the other NumericTypes.
Author: Mike Dusenberry <dusenberrymw@gmail.com>
Closes#6540 from dusenberrymw/SPARK-7426_AttributeFactory.fromStructField_Should_Allow_NumericTypes and squashes the following commits:
87fecb3 [Mike Dusenberry] Updated Attribute.fromStructField to allow any NumericType, rather than just DoubleType, and added unit tests for a few of the other NumericTypes.
Python API for PCA and PCAModel
Author: Yanbo Liang <ybliang8@gmail.com>
Closes#6315 from yanboliang/spark-7604 and squashes the following commits:
1d58734 [Yanbo Liang] remove transform() in PCAModel, use default behavior
4d9d121 [Yanbo Liang] Python API for PCA and PCAModel
JIRA: https://issues.apache.org/jira/browse/SPARK-8468
Author: Liang-Chi Hsieh <viirya@gmail.com>
Closes#6905 from viirya/cv_min and squashes the following commits:
930d3db [Liang-Chi Hsieh] Fix python unit test and add document.
d632135 [Liang-Chi Hsieh] Merge remote-tracking branch 'upstream/master' into cv_min
16e3b2c [Liang-Chi Hsieh] Take the negative instead of reciprocal.
c3dd8d9 [Liang-Chi Hsieh] For comments.
b5f52c1 [Liang-Chi Hsieh] Add param to CrossValidator for choosing whether to maximize evaulation value.
Python bindings for StreamingKMeans
Will change status to MRG once docs, tests and examples are updated.
Author: MechCoder <manojkumarsivaraj334@gmail.com>
Closes#6499 from MechCoder/spark-4118 and squashes the following commits:
7722d16 [MechCoder] minor style fixes
51052d3 [MechCoder] Doc fixes
2061a76 [MechCoder] Add tests for simultaneous training and prediction Minor style fixes
81482fd [MechCoder] minor
5d9fe61 [MechCoder] predictOn should take into account the latest model
8ab9e89 [MechCoder] Fix Python3 error
a9817df [MechCoder] Better tests and minor fixes
c80e451 [MechCoder] Add ignore_unicode_prefix
ee8ce16 [MechCoder] Update tests, doc and examples
4b1481f [MechCoder] Some changes and tests
d8b066a [MechCoder] [SPARK-4118] [MLlib] [PySpark] Python bindings for StreamingKMeans
Python API for org.apache.spark.mllib.feature.ElementwiseProduct
Author: MechCoder <manojkumarsivaraj334@gmail.com>
Closes#6346 from MechCoder/spark-7605 and squashes the following commits:
79d1ef5 [MechCoder] Consistent and support list / array types
5f81d81 [MechCoder] [SPARK-7605] [MLlib] Python API for ElementwiseProduct
MatrixUDT was recently coded in scala. This has been ported to PySpark
Author: MechCoder <manojkumarsivaraj334@gmail.com>
Closes#6354 from MechCoder/spark-6390 and squashes the following commits:
fc4dc1e [MechCoder] Better error message
c940a44 [MechCoder] Added test
aa9c391 [MechCoder] Add pyUDT to MatrixUDT
62a2a7d [MechCoder] [SPARK-6390] Port MatrixUDT to PySpark
Check then make the MLlib Python classification and regression doc to be as complete as the Scala doc.
Author: Yanbo Liang <ybliang8@gmail.com>
Closes#6460 from yanboliang/spark-7916 and squashes the following commits:
f8deda4 [Yanbo Liang] trigger jenkins
6dc4d99 [Yanbo Liang] address comments
ce2a43e [Yanbo Liang] truncate too long line and remove extra sparse
3eaf6ad [Yanbo Liang] MLlib Python doc parity check for classification and regression
MLUtils.appendBias method is heavily used in creating intercepts for linear models.
This method uses Breeze's vector concatenation which is very slow compared to the plain
System.arrayCopy. This improvement is to change the implementation to use System.arrayCopy.
I saw the following performance improvements after the change:
Benchmark with mnist dataset for 50 times:
MLUtils.appendBias (SparseVector Before): 47320 ms
MLUtils.appendBias (SparseVector After): 1935 ms
MLUtils.appendBias (DenseVector Before): 5340 ms
MLUtils.appendBias (DenseVector After): 4080 ms
This is almost a 24 times performance boost for SparseVectors.
Author: Roger Menezes <rmenezes@netflix.com>
Closes#6768 from rogermenezes/improve-append-bias and squashes the following commits:
4e42f75 [Roger Menezes] address feedback
e999d79 [Roger Menezes] first commit
Test cases for both StreamingLinearRegression and StreamingLogisticRegression, and code fix.
Edit:
This contribution is my original work and I license the work to the project under the project's open source license.
Author: Paavo <pparkkin@gmail.com>
Closes#6713 from pparkkin/streamingmodel-empty-rdd and squashes the following commits:
ff5cd78 [Paavo] Update strings to use interpolation.
db234cf [Paavo] Use !rdd.isEmpty.
54ad89e [Paavo] Test case for empty stream.
393e36f [Paavo] Ignore empty RDDs.
0bfc365 [Paavo] Test case for empty stream.
Author: MechCoder <manojkumarsivaraj334@gmail.com>
Closes#6720 from MechCoder/empty_model_check and squashes the following commits:
3a07de5 [MechCoder] Remove construct to get weights in StreamingLinearAlgorithm
This makes the constructor callable in Python. dbtsai
Author: Xiangrui Meng <meng@databricks.com>
Closes#6709 from mengxr/SPARK-8168 and squashes the following commits:
f871de4 [Xiangrui Meng] Add Python friendly constructor to PipelineModel
1. Prevent creating a map of data to find numFeatures
2. If model is empty, then initialize with a zero vector of numFeature
Author: MechCoder <manojkumarsivaraj334@gmail.com>
Closes#6684 from MechCoder/spark-8140 and squashes the following commits:
7fbf5f9 [MechCoder] [SPARK-8140] Remove empty model check in StreamingLinearAlgorithm And other minor cosmits
Python API for KernelDensity
Author: MechCoder <manojkumarsivaraj334@gmail.com>
Closes#6387 from MechCoder/spark-7639 and squashes the following commits:
17abc62 [MechCoder] add tests
2de6540 [MechCoder] style tests
bf4acc0 [MechCoder] Added doctests
84359d5 [MechCoder] [SPARK-7639] Python API for KernelDensity
Added stats from cross validation as a val in the cross validation model to save them for user access.
Author: leahmcguire <lmcguire@salesforce.com>
Closes#5915 from leahmcguire/saveCVmetrics and squashes the following commits:
49b507b [leahmcguire] fixed tyle error
67537b1 [leahmcguire] rebased
85907f0 [leahmcguire] fixed name
59987cc [leahmcguire] changed param name and test according to comments
36e71e3 [leahmcguire] rebasing
4b8223e [leahmcguire] fixed name
4ddffc6 [leahmcguire] changed param name and test according to comments
3a995da [leahmcguire] Added stats from cross validation as a val in the cross validation model to save them for user access
This is just a workaround to a bigger problem. Some pipeline stages may not be effective during prediction, and they should not complain about missing required columns, e.g. `StringIndexerModel`. jkbradley
Author: Xiangrui Meng <meng@databricks.com>
Closes#6595 from mengxr/SPARK-8051 and squashes the following commits:
b6a36b9 [Xiangrui Meng] add doc
f143fd4 [Xiangrui Meng] Merge remote-tracking branch 'apache/master' into SPARK-8051
8ee7c7e [Xiangrui Meng] use SparkFunSuite
e112394 [Xiangrui Meng] make StringIndexerModel silent if input column does not exist
Java-friendly APIs added:
* GaussianMixture.run()
* GaussianMixtureModel.predict()
* DistributedLDAModel.javaTopicDistributions()
* StreamingKMeans: trainOn, predictOn, predictOnValues
* Statistics.corr
* params
* added doc to w() since Java docs do not inherit doc
* removed non-Java-friendly w() from StringArrayParam and DoubleArrayParam
* made DoubleArrayParam Java-friendly w() actually Java-friendly
I generated the doc and verified all changes.
CC: mengxr
Author: Joseph K. Bradley <joseph@databricks.com>
Closes#6562 from jkbradley/java-api-1.4 and squashes the following commits:
c16821b [Joseph K. Bradley] Small fixes based on code review.
d955581 [Joseph K. Bradley] unit test fixes
29b6b0d [Joseph K. Bradley] small fixes
fe6dcfe [Joseph K. Bradley] Added several Java-friendly APIs + unit tests: NaiveBayes, GaussianMixture, LDA, StreamingKMeans, Statistics.corr, params
Author: Patrick Wendell <patrick@databricks.com>
Closes#6328 from pwendell/spark-1.5-update and squashes the following commits:
2f42d02 [Patrick Wendell] A few more excludes
4bebcf0 [Patrick Wendell] Update to RC4
61aaf46 [Patrick Wendell] Using new release candidate
55f1610 [Patrick Wendell] Another exclude
04b4f04 [Patrick Wendell] More issues with transient 1.4 changes
36f549b [Patrick Wendell] [SPARK-7801] [BUILD] Updating versions to SPARK 1.5.0
jira: https://issues.apache.org/jira/browse/SPARK-7983
Customers frequently use zero-based indices in their LIBSVM files. No warnings or errors from Spark will be reported during their computation afterwards, and usually it will lead to wired result for many algorithms (like GBDT).
add a quick check.
Author: Yuhao Yang <hhbyyh@gmail.com>
Closes#6538 from hhbyyh/loadSVM and squashes the following commits:
79d9c11 [Yuhao Yang] optimization as respond to comments
4310710 [Yuhao Yang] merge conflict
96460f1 [Yuhao Yang] merge conflict
20a2811 [Yuhao Yang] use require
6e4f8ca [Yuhao Yang] add check for ascending order
9956365 [Yuhao Yang] add ut for 0-based loadlibsvm exception
5bd1f9a [Yuhao Yang] add require for one-based in loadLIBSVM
I searched the Spark codebase for all occurrences of "scalingVector"
CC: mengxr
Author: Joseph K. Bradley <joseph@databricks.com>
Closes#6596 from jkbradley/scalingVec-rename and squashes the following commits:
d3812f8 [Joseph K. Bradley] renamed scalingVector to scalingVec
This patch significantly refactors CatalystTypeConverters to both clean up the code and enable these conversions to work with future Project Tungsten features.
At a high level, I've reorganized the code so that all functions dealing with the same type are grouped together into type-specific subclasses of `CatalystTypeConveter`. In addition, I've added new methods that allow the Catalyst Row -> Scala Row conversions to access the Catalyst row's fields through type-specific `getTYPE()` methods rather than the generic `get()` / `Row.apply` methods. This refactoring is a blocker to being able to unit test new operators that I'm developing as part of Project Tungsten, since those operators may output `UnsafeRow` instances which don't support the generic `get()`.
The stricter type usage of types here has uncovered some bugs in other parts of Spark SQL:
- #6217: DescribeCommand is assigned wrong output attributes in SparkStrategies
- #6218: DataFrame.describe() should cast all aggregates to String
- #6400: Use output schema, not relation schema, for data source input conversion
Spark SQL current has undefined behavior for what happens when you try to create a DataFrame from user-specified rows whose values don't match the declared schema. According to the `createDataFrame()` Scaladoc:
> It is important to make sure that the structure of every [[Row]] of the provided RDD matches the provided schema. Otherwise, there will be runtime exception.
Given this, it sounds like it's technically not a break of our API contract to fail-fast when the data types don't match. However, there appear to be many cases where we don't fail even though the types don't match. For example, `JavaHashingTFSuite.hasingTF` passes a column of integers values for a "label" column which is supposed to contain floats. This column isn't actually read or modified as part of query processing, so its actual concrete type doesn't seem to matter. In other cases, there could be situations where we have generic numeric aggregates that tolerate being called with different numeric types than the schema specified, but this can be okay due to numeric conversions.
In the long run, we will probably want to come up with precise semantics for implicit type conversions / widening when converting Java / Scala rows to Catalyst rows. Until then, though, I think that failing fast with a ClassCastException is a reasonable behavior; this is the approach taken in this patch. Note that certain optimizations in the inbound conversion functions for primitive types mean that we'll probably preserve the old undefined behavior in a majority of cases.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#6222 from JoshRosen/catalyst-converters-refactoring and squashes the following commits:
740341b [Josh Rosen] Optimize method dispatch for primitive type conversions
befc613 [Josh Rosen] Add tests to document Option-handling behavior.
5989593 [Josh Rosen] Use new SparkFunSuite base in CatalystTypeConvertersSuite
6edf7f8 [Josh Rosen] Re-add convertToScala(), since a Hive test still needs it
3f7b2d8 [Josh Rosen] Initialize converters lazily so that the attributes are resolved first
6ad0ebb [Josh Rosen] Fix JavaHashingTFSuite ClassCastException
677ff27 [Josh Rosen] Fix null handling bug; add tests.
8033d4c [Josh Rosen] Fix serialization error in UserDefinedGenerator.
85bba9d [Josh Rosen] Fix wrong input data in InMemoryColumnarQuerySuite
9c0e4e1 [Josh Rosen] Remove last use of convertToScala().
ae3278d [Josh Rosen] Throw ClassCastException errors during inbound conversions.
7ca7fcb [Josh Rosen] Comments and cleanup
1e87a45 [Josh Rosen] WIP refactoring of CatalystTypeConverters
This is scala example code for both linear and logistic regression. Python and Java versions are to be added.
Author: DB Tsai <dbt@netflix.com>
Closes#6576 from dbtsai/elasticNetExample and squashes the following commits:
e7ca406 [DB Tsai] fix test
6bb6d77 [DB Tsai] fix suite and remove duplicated setMaxIter
136e0dd [DB Tsai] address feedback
1ec29d4 [DB Tsai] fix style
9462f5f [DB Tsai] add example
The temporary column should be dropped after we get the prediction column. harsha2010
Author: Xiangrui Meng <meng@databricks.com>
Closes#6592 from mengxr/SPARK-8049 and squashes the following commits:
1d89107 [Xiangrui Meng] use SparkFunSuite
6ee70de [Xiangrui Meng] drop tmp col from OneVsRest output
Updating ML Doc's *"Estimator, Transformer, and Param"* example to use `model.extractParamMap` instead of `model.fittingParamMap`, which no longer exists.
mengxr, I believe this addresses (part of) the *update documentation* TODO list item from [PR 5820](https://github.com/apache/spark/pull/5820).
Author: Mike Dusenberry <dusenberrymw@gmail.com>
Closes#6514 from dusenberrymw/Fix_ML_Doc_Estimator_Transformer_Param_Example and squashes the following commits:
6366e1f [Mike Dusenberry] Updating instances of model.extractParamMap to model.parent.extractParamMap, since the Params of the parent Estimator could possibly differ from thos of the Model.
d850e0e [Mike Dusenberry] Removing all references to "fittingParamMap" throughout Spark, since it has been removed.
0480304 [Mike Dusenberry] Updating the ML Doc "Estimator, Transformer, and Param" Java example to use model.extractParamMap() instead of model.fittingParamMap(), which no longer exists.
7d34939 [Mike Dusenberry] Updating ML Doc "Estimator, Transformer, and Param" example to use model.extractParamMap instead of model.fittingParamMap, which no longer exists.
This PR adds a Java unit test and user guide for `StringIndexer`. I put it before `OneHotEncoder` because they are closely related. jkbradley
Author: Xiangrui Meng <meng@databricks.com>
Closes#6561 from mengxr/SPARK-7582 and squashes the following commits:
4bba4f1 [Xiangrui Meng] fix example
ba1cd1b [Xiangrui Meng] fix style
7fa18d1 [Xiangrui Meng] add user guide for StringIndexer
136cb93 [Xiangrui Meng] add a Java unit test for StringIndexer
This PR adds a section in the user guide for `VectorAssembler` with code examples in Python/Java/Scala. It also adds a unit test in Java.
jkbradley
Author: Xiangrui Meng <meng@databricks.com>
Closes#6556 from mengxr/SPARK-7584 and squashes the following commits:
11313f6 [Xiangrui Meng] simplify Java example
0cd47f3 [Xiangrui Meng] update user guide
fd36292 [Xiangrui Meng] update Java unit test
ce61ca0 [Xiangrui Meng] add Java unit test for VectorAssembler
e399942 [Xiangrui Meng] scala/python example code
Author: Reynold Xin <rxin@databricks.com>
Closes#6534 from rxin/whitespace-mllib and squashes the following commits:
38926e3 [Reynold Xin] [SPARK-3850] Trim trailing spaces for MLlib.
The MLlib ChiSqSelector class is not serializable, and so the example in the ChiSqSelector documentation fails. Also, that example is missing the import of ChiSqSelector.
This PR makes ChiSqSelector extend Serializable in MLlib, and adds the ChiSqSelector import statement to the associated example in the documentation.
Author: Mike Dusenberry <dusenberrymw@gmail.com>
Closes#6462 from dusenberrymw/Make_ChiSqSelector_Serializable_and_Fix_Related_Docs_Example and squashes the following commits:
9cb2f94 [Mike Dusenberry] Make MLlib ChiSqSelector Serializable.
d9003bf [Mike Dusenberry] Add missing import in MLlib ChiSqSelector Docs Scala example.
This patch fixes a build break in maven caused by #6441.
Note that this patch reverts the changes in flume-sink because
this module does not currently depend on Spark core, but the
tests require it. There is not an easy way to make this work
because mvn test dependencies are not transitive (MNG-1378).
For now, we will leave the one test suite in flume-sink out
until we figure out a better solution. This patch is mainly
intended to unbreak the maven build.
Author: Andrew Or <andrew@databricks.com>
Closes#6511 from andrewor14/fix-build-mvn and squashes the following commits:
3d53643 [Andrew Or] [HOT FIX#6441] Fix maven build failures
Right now `unit-tests.log` are not of much value because we can't tell where the test boundaries are easily. This patch adds log statements before and after each test to outline the test boundaries, e.g.:
```
===== TEST OUTPUT FOR o.a.s.serializer.KryoSerializerSuite: 'kryo with parallelize for primitive arrays' =====
15/05/27 12:36:39.596 pool-1-thread-1-ScalaTest-running-KryoSerializerSuite INFO SparkContext: Starting job: count at KryoSerializerSuite.scala:230
15/05/27 12:36:39.596 dag-scheduler-event-loop INFO DAGScheduler: Got job 3 (count at KryoSerializerSuite.scala:230) with 4 output partitions (allowLocal=false)
15/05/27 12:36:39.596 dag-scheduler-event-loop INFO DAGScheduler: Final stage: ResultStage 3(count at KryoSerializerSuite.scala:230)
15/05/27 12:36:39.596 dag-scheduler-event-loop INFO DAGScheduler: Parents of final stage: List()
15/05/27 12:36:39.597 dag-scheduler-event-loop INFO DAGScheduler: Missing parents: List()
15/05/27 12:36:39.597 dag-scheduler-event-loop INFO DAGScheduler: Submitting ResultStage 3 (ParallelCollectionRDD[5] at parallelize at KryoSerializerSuite.scala:230), which has no missing parents
...
15/05/27 12:36:39.624 pool-1-thread-1-ScalaTest-running-KryoSerializerSuite INFO DAGScheduler: Job 3 finished: count at KryoSerializerSuite.scala:230, took 0.028563 s
15/05/27 12:36:39.625 pool-1-thread-1-ScalaTest-running-KryoSerializerSuite INFO KryoSerializerSuite:
***** FINISHED o.a.s.serializer.KryoSerializerSuite: 'kryo with parallelize for primitive arrays' *****
...
```
Author: Andrew Or <andrew@databricks.com>
Closes#6441 from andrewor14/demarcate-tests and squashes the following commits:
879b060 [Andrew Or] Fix compile after rebase
d622af7 [Andrew Or] Merge branch 'master' of github.com:apache/spark into demarcate-tests
017c8ba [Andrew Or] Merge branch 'master' of github.com:apache/spark into demarcate-tests
7790b6c [Andrew Or] Fix tests after logical merge conflict
c7460c0 [Andrew Or] Merge branch 'master' of github.com:apache/spark into demarcate-tests
c43ffc4 [Andrew Or] Fix tests?
8882581 [Andrew Or] Fix tests
ee22cda [Andrew Or] Fix log message
fa9450e [Andrew Or] Merge branch 'master' of github.com:apache/spark into demarcate-tests
12d1e1b [Andrew Or] Various whitespace changes (minor)
69cbb24 [Andrew Or] Make all test suites extend SparkFunSuite instead of FunSuite
bbce12e [Andrew Or] Fix manual things that cannot be covered through automation
da0b12f [Andrew Or] Add core tests as dependencies in all modules
f7d29ce [Andrew Or] Introduce base abstract class for all test suites
Author: MechCoder <manojkumarsivaraj334@gmail.com>
Closes#6497 from MechCoder/spark-7946 and squashes the following commits:
2fdd0a3 [MechCoder] Add non-regression test
8c988c6 [MechCoder] [SPARK-7946] DecayFactor wrongly set in StreamingKMeans
This PR contains two major changes to `OneHotEncoder`:
1. more robust handling of ML attributes. If the input attribute is unknown, we look at the values to get the max category index
2. change `includeFirst` to `dropLast` and leave the default to `true`. There are couple benefits:
a. consistent with other tutorials of one-hot encoding (or dummy coding) (e.g., http://www.ats.ucla.edu/stat/mult_pkg/faq/general/dummy.htm)
b. keep the indices unmodified in the output vector. If we drop the first, all indices will be shifted by 1.
c. If users use `StringIndex`, the last element is the least frequent one.
Sorry for including two changes in one PR! I'll update the user guide in another PR.
jkbradley sryza
Author: Xiangrui Meng <meng@databricks.com>
Closes#6466 from mengxr/SPARK-7912 and squashes the following commits:
a280dca [Xiangrui Meng] fix tests
d8f234d [Xiangrui Meng] Merge remote-tracking branch 'apache/master' into SPARK-7912
171b276 [Xiangrui Meng] mention the difference between our impl vs sklearn's
00dfd96 [Xiangrui Meng] update OneHotEncoder in Python
208ddad [Xiangrui Meng] update OneHotEncoder to handle ML attributes and change includeFirst to dropLast
Expose user/item factors in DataFrames. This is to be more consistent with the pipeline API. It also helps maintain consistent APIs across languages. This PR also removed fitting params from `ALSModel`.
coderxiang
Author: Xiangrui Meng <meng@databricks.com>
Closes#6468 from mengxr/SPARK-7922 and squashes the following commits:
7bfb1d5 [Xiangrui Meng] update ALSModel in PySpark
1ba5607 [Xiangrui Meng] use DataFrames for user/item factors in ALS
`VectorAssembler` should carry over ML attributes. For unknown attributes, we assume numeric values. This PR handles the following cases:
1. DoubleType with ML attribute: carry over
2. DoubleType without ML attribute: numeric value
3. Scalar type: numeric value
4. VectorType with all ML attributes: carry over and update names
5. VectorType with number of ML attributes: assume all numeric
6. VectorType without ML attributes: check the first row and get the number of attributes
jkbradley
Author: Xiangrui Meng <meng@databricks.com>
Closes#6452 from mengxr/SPARK-7198 and squashes the following commits:
a9d2469 [Xiangrui Meng] add space
facdb1f [Xiangrui Meng] VectorAssembler should output ML attributes
~~A PythonUDT shouldn't be serialized into external Scala types in PythonRDD. I'm not sure whether this should fix one of the bugs related to SQL UDT/UDF in PySpark.~~
The fix above didn't work. So I added a workaround for this. If a Python UDF is applied to a Python UDT. This will put the Python SQL types as inputs. Still incorrect, but at least it doesn't throw exceptions on the Scala side. davies harsha2010
Author: Xiangrui Meng <meng@databricks.com>
Closes#6442 from mengxr/SPARK-7903 and squashes the following commits:
c257d2a [Xiangrui Meng] add a workaround for VectorUDT
With descent coverage of feature transformers, algorithms, and model tuning support, it is time to graduate `spark.ml` from alpha. This PR changes all `AlphaComponent` annotations to either `DeveloperApi` or `Experimental`, depending on whether we expect a class/method to be used by end users (who use the pipeline API to assemble/tune their ML pipelines but not to create new pipeline components.) `UnaryTransformer` becomes a `DeveloperApi` in this PR.
jkbradley harsha2010
Author: Xiangrui Meng <meng@databricks.com>
Closes#6417 from mengxr/SPARK-7748 and squashes the following commits:
effbccd [Xiangrui Meng] organize imports
c15028e [Xiangrui Meng] added missing docs
1b2e5f8 [Xiangrui Meng] update package doc
73ca791 [Xiangrui Meng] alpha -> ex/dev for the rest
93819db [Xiangrui Meng] alpha -> ex/dev in ml.param
55ca073 [Xiangrui Meng] alpha -> ex/dev in ml.feature
83572f1 [Xiangrui Meng] add Experimental and DeveloperApi tags (wip)
The densities in KernelDensity are scaled down by
(number of parallel processes X number of points). It should be just no.of samples. This results in broken tests in KernelDensitySuite which haven't been tested properly.
Author: MechCoder <manojkumarsivaraj334@gmail.com>
Closes#6383 from MechCoder/spark-7844 and squashes the following commits:
ab81302 [MechCoder] Math->math
9b8ed50 [MechCoder] Make one pass to update count
a92fe50 [MechCoder] [SPARK-7844] Fix broken tests in KernelDensity
Added user guide sections with code examples.
Also added small Java unit tests to test Java example in guide.
CC: mengxr
Author: Joseph K. Bradley <joseph@databricks.com>
Closes#6127 from jkbradley/feature-guide-2 and squashes the following commits:
cd47f4b [Joseph K. Bradley] Updated based on code review
f16bcec [Joseph K. Bradley] Fixed merge issues and update Python examples print calls for Python 3
0a862f9 [Joseph K. Bradley] Added Normalizer, StandardScaler to ml-features doc, plus small Java unit tests
a21c2d6 [Joseph K. Bradley] Updated ml-features.md with IDF
This PR updates `HashingTF` to output ML attributes that tell the number of features in the output column. We need to expand `UnaryTransformer` to support output metadata. A `df outputMetadata: Metadata` is not sufficient because the metadata may also depends on the input data. Though this is not true for `HashingTF`, I think it is reasonable to update `UnaryTransformer` in a separate PR. `checkParams` is added to verify common requirements for params. I will send a separate PR to use it in other test suites. jkbradley
Author: Xiangrui Meng <meng@databricks.com>
Closes#6308 from mengxr/SPARK-7219 and squashes the following commits:
9bd2922 [Xiangrui Meng] address comments
e82a68a [Xiangrui Meng] remove sqlContext from test suite
995535b [Xiangrui Meng] Merge remote-tracking branch 'apache/master' into SPARK-7219
2194703 [Xiangrui Meng] add test for attributes
178ae23 [Xiangrui Meng] update HashingTF with tests
91a6106 [Xiangrui Meng] WIP
The previous default is `{gaps: false, pattern: "\\p{L}+|[^\\p{L}\\s]+"}`. The default pattern is hard to understand. This PR changes the default to `{gaps: true, pattern: "\\s+"}`. jkbradley
Author: Xiangrui Meng <meng@databricks.com>
Closes#6330 from mengxr/SPARK-7794 and squashes the following commits:
5ee7cde [Xiangrui Meng] update RegexTokenizer default settings
We removed `varargs` due to Java compilation issues. That was a false alarm because I didn't run `build/sbt clean`. So this PR reverts the changes. jkbradley
Author: Xiangrui Meng <meng@databricks.com>
Closes#6320 from mengxr/SPARK-7498 and squashes the following commits:
74a7259 [Xiangrui Meng] add varargs back to setDefault
Added VectorIndexer section to ML user guide. Also added javaCategoryMaps() method and Java unit test for it.
CC: mengxr
Author: Joseph K. Bradley <joseph@databricks.com>
Closes#6255 from jkbradley/vector-indexer-guide and squashes the following commits:
dbb8c4c [Joseph K. Bradley] simplified VectorIndexerModel.javaCategoryMaps
f692084 [Joseph K. Bradley] Added VectorIndexer section to ML user guide. Also added javaCategoryMaps() method and Java unit test for it.
same issue and fix as in Spark-7694.
Author: Shuo Xiang <shuoxiangpub@gmail.com>
Closes#6321 from coderxiang/nb and squashes the following commits:
a5e6de4 [Shuo Xiang] use getOrElse for svmmodel.tostring
2cb0177 [Shuo Xiang] Merge remote-tracking branch 'upstream/master' into nb
5f109b4 [Shuo Xiang] Merge remote-tracking branch 'upstream/master'
c5c5bfe [Shuo Xiang] Merge remote-tracking branch 'upstream/master'
98804c9 [Shuo Xiang] fix bug in topBykey and update test
to be consistent with other string names in MLlib. This PR also updates the implementation to use vals instead of hardcoded strings. jkbradley leahmcguire
Author: Xiangrui Meng <meng@databricks.com>
Closes#6277 from mengxr/SPARK-7752 and squashes the following commits:
f38b662 [Xiangrui Meng] add another case _ back in test
ae5c66a [Xiangrui Meng] model type -> modelType
711d1c6 [Xiangrui Meng] Merge remote-tracking branch 'apache/master' into SPARK-7752
40ae53e [Xiangrui Meng] fix Java test suite
264a814 [Xiangrui Meng] add case _ back
3c456a8 [Xiangrui Meng] update NB user guide
17bba53 [Xiangrui Meng] update naive Bayes to use lowercase model type strings
Update `KernelDensity` API to make it extensible to different kernels in the future. `bandwidth` is used instead of `standardDeviation`. The static `kernelDensity` method is removed from `Statistics`. The implementation is updated using BLAS, while the algorithm remains the same. sryza srowen
Author: Xiangrui Meng <meng@databricks.com>
Closes#6279 from mengxr/SPARK-7753 and squashes the following commits:
4cdfadc [Xiangrui Meng] add example code in the doc
767fd5a [Xiangrui Meng] update KernelDensity API
to simplify test suites that require a SQLContext.
Author: Xiangrui Meng <meng@databricks.com>
Closes#6303 from mengxr/SPARK-7774 and squashes the following commits:
0622b5a [Xiangrui Meng] update some other test suites
e1f9b8d [Xiangrui Meng] add sqlContext to MLlibTestSparkContext
Set a default value for `outputCol` instead of forcing users to name it. This is useful for intermediate transformers in the pipeline. jkbradley
Author: Xiangrui Meng <meng@databricks.com>
Closes#6289 from mengxr/SPARK-7762 and squashes the following commits:
54edebc [Xiangrui Meng] merge master
bff8667 [Xiangrui Meng] update unit test
171246b [Xiangrui Meng] add unit test for outputCol
a4321bd [Xiangrui Meng] set default value for outputCol
Minor updates to the spark.mllib APIs:
1. Add `DeveloperApi` to `PMMLExportable` and add `Experimental` to `toPMML` methods.
2. Mention `RankingMetrics.of` in the `RankingMetrics` constructor.
Author: Xiangrui Meng <meng@databricks.com>
Closes#6280 from mengxr/SPARK-7537 and squashes the following commits:
1bd2583 [Xiangrui Meng] organize imports
94afa7a [Xiangrui Meng] mark all toPMML methods experimental
4c40da1 [Xiangrui Meng] mention the factory method for RankingMetrics for Java users
88c62d0 [Xiangrui Meng] add DeveloperApi to PMMLExportable
Add MultilabelMetrics in PySpark/MLlib
Author: Yanbo Liang <ybliang8@gmail.com>
Closes#6276 from yanboliang/spark-6094 and squashes the following commits:
b8e3343 [Yanbo Liang] Add MultilabelMetrics in PySpark/MLlib
JIRA issue [link](https://issues.apache.org/jira/browse/SPARK-7663).
We should check the model size of word2vec, to prevent the unexpected empty.
CC srowen.
Author: Xusen Yin <yinxusen@gmail.com>
Closes#6228 from yinxusen/SPARK-7663 and squashes the following commits:
21770c5 [Xusen Yin] check the vocab size
54ae63e [Xusen Yin] add requirement for word2vec model
CC jkbradley.
JIRA [issue](https://issues.apache.org/jira/browse/SPARK-7586).
Author: Xusen Yin <yinxusen@gmail.com>
Closes#6181 from yinxusen/SPARK-7586 and squashes the following commits:
77014c5 [Xusen Yin] comment fix
57a4c07 [Xusen Yin] small fix for docs
1178c8f [Xusen Yin] remove the correctness check in java suite
1c3f389 [Xusen Yin] delete sbt commit
1af152b [Xusen Yin] check python example code
1b5369e [Xusen Yin] add docs of word2vec
Changed shared param HasSeed to have default based on hashCode of class name, instead of random number.
Also, removed fixed random seeds from Word2Vec and ALS.
CC: mengxr
Author: Joseph K. Bradley <joseph@databricks.com>
Closes#6251 from jkbradley/scala-fixed-seed and squashes the following commits:
0e37184 [Joseph K. Bradley] Fixed Word2VecSuite, ALSSuite in spark.ml to use original fixed random seeds
678ec3a [Joseph K. Bradley] Removed fixed random seeds from Word2Vec and ALS. Changed shared param HasSeed to have default based on hashCode of class name, instead of random number.
Made Model.parent transient. Added Model.hasParent to test for null parent
CC: mengxr
Author: Joseph K. Bradley <joseph@databricks.com>
Closes#5914 from jkbradley/parent-optional and squashes the following commits:
d501774 [Joseph K. Bradley] Made Model.parent transient. Added Model.hasParent to test for null parent
JIRA [here](https://issues.apache.org/jira/browse/SPARK-7581).
CC jkbradley
Author: Xusen Yin <yinxusen@gmail.com>
Closes#6113 from yinxusen/SPARK-7581 and squashes the following commits:
1a7d80d [Xusen Yin] merge with master
892a8e9 [Xusen Yin] fix python 3 compatibility
ec935bf [Xusen Yin] small fix
3e9fa1d [Xusen Yin] delete note
69fcf85 [Xusen Yin] simplify and add python example
81d21dc [Xusen Yin] add programming guide for Polynomial Expansion
40babfb [Xusen Yin] add java test suite for PolynomialExpansion
JIRA: https://issues.apache.org/jira/browse/SPARK-7681
Author: Liang-Chi Hsieh <viirya@gmail.com>
Closes#6209 from viirya/sparsevector_gemv and squashes the following commits:
ce0bb8b [Liang-Chi Hsieh] Still need to scal y when beta is 0.0 because it clears out y.
b890e63 [Liang-Chi Hsieh] Do not delete multiply for DenseVector.
57a8c1e [Liang-Chi Hsieh] Add MimaExcludes for v1.4.
458d1ae [Liang-Chi Hsieh] List DenseMatrix.multiply and SparseMatrix.multiply to MimaExcludes too.
054f05d [Liang-Chi Hsieh] Fix scala style.
410381a [Liang-Chi Hsieh] Address comments. Make Matrix.multiply more generalized.
4616696 [Liang-Chi Hsieh] Add support for SparseVector with SparseMatrix.
5d6d07a [Liang-Chi Hsieh] Merge remote-tracking branch 'upstream/master' into sparsevector_gemv
c069507 [Liang-Chi Hsieh] Add SparseVector support for gemv with DenseMatrix.
This PR makes pipeline stages in Python copyable and hence simplifies some implementations. It also includes the following changes:
1. Rename `paramMap` and `defaultParamMap` to `_paramMap` and `_defaultParamMap`, respectively.
2. Accept a list of param maps in `fit`.
3. Use parent uid and name to identify param.
jkbradley
Author: Xiangrui Meng <meng@databricks.com>
Author: Joseph K. Bradley <joseph@databricks.com>
Closes#6088 from mengxr/SPARK-7380 and squashes the following commits:
413c463 [Xiangrui Meng] remove unnecessary doc
4159f35 [Xiangrui Meng] Merge remote-tracking branch 'apache/master' into SPARK-7380
611c719 [Xiangrui Meng] fix python style
68862b8 [Xiangrui Meng] update _java_obj initialization
927ad19 [Xiangrui Meng] fix ml/tests.py
0138fc3 [Xiangrui Meng] update feature transformers and fix a bug in RegexTokenizer
9ca44fb [Xiangrui Meng] simplify Java wrappers and add tests
c7d84ef [Xiangrui Meng] update ml/tests.py to test copy params
7e0d27f [Xiangrui Meng] merge master
46840fb [Xiangrui Meng] update wrappers
b6db1ed [Xiangrui Meng] update all self.paramMap to self._paramMap
46cb6ed [Xiangrui Meng] merge master
a163413 [Xiangrui Meng] fix style
1042e80 [Xiangrui Meng] Merge remote-tracking branch 'apache/master' into SPARK-7380
9630eae [Xiangrui Meng] fix Identifiable._randomUID
13bd70a [Xiangrui Meng] update ml/tests.py
64a536c [Xiangrui Meng] use _fit/_transform/_evaluate to simplify the impl
02abf13 [Xiangrui Meng] Merge remote-tracking branch 'apache/master' into copyable-python
66ce18c [Joseph K. Bradley] some cleanups before sending to Xiangrui
7431272 [Joseph K. Bradley] Rebased with master
The `toString` method of `LogisticRegressionModel` calls `get` method on an Option (threshold) without a safeguard. In spark-shell, the following code `val model = algorithm.run(data).clearThreshold()` in lbfgs code will fail as `toString `method will be called right after `clearThreshold()` to show the results in the REPL.
Author: Shuo Xiang <shuoxiangpub@gmail.com>
Closes#6224 from coderxiang/getorelse and squashes the following commits:
d5f53c9 [Shuo Xiang] use getOrElse for getting the threshold of LR model
5f109b4 [Shuo Xiang] Merge remote-tracking branch 'upstream/master'
c5c5bfe [Shuo Xiang] Merge remote-tracking branch 'upstream/master'
98804c9 [Shuo Xiang] fix bug in topBykey and update test
Author: Reynold Xin <rxin@databricks.com>
Closes#6211 from rxin/mllib-reader and squashes the following commits:
79a2cb9 [Reynold Xin] [SPARK-7654][MLlib] Migrate MLlib to the DataFrame reader/writer API.
reservoir feature sample by using existing api
Author: AiHe <ai.he@ussuning.com>
Closes#5988 from AiHe/reservoir and squashes the following commits:
e7a41ac [AiHe] remove non-robust testing case
28ffb9a [AiHe] set seed as rng.nextLong
37459e1 [AiHe] set fixed seed
1e98a4c [AiHe] [MLLIB][tree] Add reservoir sample in RandomForest
JIRA: https://issues.apache.org/jira/browse/SPARK-7668
Author: Liang-Chi Hsieh <viirya@gmail.com>
Closes#6188 from viirya/fix_matrix_map and squashes the following commits:
2a7cc97 [Liang-Chi Hsieh] Preserve isTransposed property for Matrix after calling map function.
Implement Python API for major disparities of GaussianMixture cluster algorithm between Scala & Python
```scala
GaussianMixture
setInitialModel
GaussianMixtureModel
k
```
Author: Yanbo Liang <ybliang8@gmail.com>
Closes#6087 from yanboliang/spark-6258 and squashes the following commits:
b3af21c [Yanbo Liang] fix typo
2b645c1 [Yanbo Liang] fix doc
638b4b7 [Yanbo Liang] address comments
b5bcade [Yanbo Liang] GaussianMixture Python API parity check
A param instance is strongly attached to an parent in the current implementation. So if we make a copy of an estimator or a transformer in pipelines and other meta-algorithms, it becomes error-prone to copy the params to the copied instances. In this PR, a param is identified by its parent's UID and the param name. So it becomes loosely attached to its parent and all its derivatives. The UID is preserved during copying or fitting. All components now have a default constructor and a constructor that takes a UID as input. I keep the constructors for Param in this PR to reduce the amount of diff and moved `parent` as a mutable field.
This PR still needs some clean-ups, and there are several spark.ml PRs pending. I'll try to get them merged first and then update this PR.
jkbradley
Author: Xiangrui Meng <meng@databricks.com>
Closes#6019 from mengxr/SPARK-7407 and squashes the following commits:
c4c8120 [Xiangrui Meng] Merge remote-tracking branch 'apache/master' into SPARK-7407
520f0a2 [Xiangrui Meng] address comments
2569168 [Xiangrui Meng] fix tests
873caca [Xiangrui Meng] fix tests in OneVsRest; fix a racing condition in shouldOwn
409ea08 [Xiangrui Meng] minor updates
83a163c [Xiangrui Meng] update JavaDeveloperApiExample
5db5325 [Xiangrui Meng] update OneVsRest
7bde7ae [Xiangrui Meng] merge master
697fdf9 [Xiangrui Meng] update Bucketizer
7b4f6c2 [Xiangrui Meng] Merge remote-tracking branch 'apache/master' into SPARK-7407
629d402 [Xiangrui Meng] fix LRSuite
154516f [Xiangrui Meng] merge master
aa4a611 [Xiangrui Meng] fix examples/compile
a4794dd [Xiangrui Meng] change Param to use to reduce the size of diff
fdbc415 [Xiangrui Meng] all tests passed
c255f17 [Xiangrui Meng] fix tests in ParamsSuite
818e1db [Xiangrui Meng] merge master
e1160cf [Xiangrui Meng] fix tests
fbc39f0 [Xiangrui Meng] pass test:compile
108937e [Xiangrui Meng] pass compile
8726d39 [Xiangrui Meng] use parent uid in Param
eaeed35 [Xiangrui Meng] update Identifiable
Author: DB Tsai <dbt@netflix.com>
Closes#6137 from dbtsai/clean and squashes the following commits:
185816d [DB Tsai] fix compilication issue
f418d08 [DB Tsai] first commit
This is similar to the changes to k-means, which gives us better control on the performance. dbtsai
Author: Xiangrui Meng <meng@databricks.com>
Closes#6128 from mengxr/SPARK-7612 and squashes the following commits:
b5c24c5 [Xiangrui Meng] merge master
a90e3ec [Xiangrui Meng] update NB training to use mllib's BLAS
Author: leahmcguire <lmcguire@salesforce.com>
Closes#6073 from leahmcguire/binaryCheckNB and squashes the following commits:
b8442c2 [leahmcguire] changed to if else for value checks
911bf83 [leahmcguire] undid reformat
4eedf1e [leahmcguire] moved bernoulli check
9ee9e84 [leahmcguire] fixed style error
3f3b32c [leahmcguire] fixed zero one check so only called in combiner
831fd27 [leahmcguire] got test working
f44bb3c [leahmcguire] removed changes from CV branch
67253f0 [leahmcguire] added check to bernoulli to ensure feature values are zero or one
f191c71 [leahmcguire] fixed name
58d060b [leahmcguire] changed param name and test according to comments
04f0d3c [leahmcguire] Added stats from cross validation as a val in the cross validation model to save them for user access
Added `ml.feature.Bucketizer` to PySpark.
cc mengxr
Author: Burak Yavuz <brkyvz@gmail.com>
Closes#6124 from brkyvz/ml-bucket and squashes the following commits:
05285be [Burak Yavuz] added sphinx doc
6abb6ed [Burak Yavuz] added support for Bucketizer
`RankingMetrics` contains a ClassTag, which is hard to create in Java. This PR adds a factory method `of` for Java users. coderxiang
Author: Xiangrui Meng <meng@databricks.com>
Closes#6098 from mengxr/SPARK-7528 and squashes the following commits:
e5d57ae [Xiangrui Meng] make RankingMetrics Java-friendly
Minor cleanups discussed with [~mengxr]:
* move OneVsRest from reduction to classification sub-package
* make model constructor private
Some doc cleanups too
CC: harsha2010 Could you please verify this looks OK? Thanks!
Author: Joseph K. Bradley <joseph@databricks.com>
Closes#6097 from jkbradley/onevsrest-cleanup and squashes the following commits:
4ecd48d [Joseph K. Bradley] org imports
430b065 [Joseph K. Bradley] moved OneVsRest from reduction subpackage to classification. small java doc style fixes
9f8b9b9 [Joseph K. Bradley] Small cleanups to OneVsRest. Made model constructor private to ml package.
Added feature transformer subsection to spark.ml guide, with HashingTF and Tokenizer. Added JavaHashingTFSuite to test Java examples in new guide.
I've run Scala, Python examples in the Spark/PySpark shells. I ran the Java examples via the test suite (with small modifications for printing).
CC: mengxr
Author: Joseph K. Bradley <joseph@databricks.com>
Closes#6093 from jkbradley/hashingtf-guide and squashes the following commits:
d5d213f [Joseph K. Bradley] small fix
dd6e91a [Joseph K. Bradley] fixes from code review of user guide
33c3ff9 [Joseph K. Bradley] small fix
bc6058c [Joseph K. Bradley] fix link
361a174 [Joseph K. Bradley] Added subsection for feature transformers to spark.ml guide, with HashingTF and Tokenizer. Added JavaHashingTFSuite to test Java examples in new guide
`scala.Math` is deprecated since 2.8. This PR only touchs `Math` usages in MLlib. dbtsai
Author: Xiangrui Meng <meng@databricks.com>
Closes#6092 from mengxr/SPARK-7571 and squashes the following commits:
fe8f8d3 [Xiangrui Meng] Math -> math
We make special treatment for +inf in `Bucketizer`. This could be simplified by always including the largest split value in the last bucket. E.g., (x1, x2, x3) defines buckets [x1, x2) and [x2, x3]. This shouldn't affect user code much, and there are applications that need to include the right-most value. For example, we can bucketize ratings from 0 to 10 to bad, neutral, and good with splits 0, 4, 6, 10. It may reads weird if the users need to put 0, 4, 6, 10.1 (or 11).
This also update the impl to use `Arrays.binarySearch` and `withClue` in test.
yinxusen jkbradley
Author: Xiangrui Meng <meng@databricks.com>
Closes#6075 from mengxr/SPARK-7559 and squashes the following commits:
e28f910 [Xiangrui Meng] update bucketizer impl
The sbt part of the build is hacky; it basically tricks sbt
into generating the zip by using a generator, but returns
an empty list for the generated files so that nothing is
actually added to the assembly.
Author: Marcelo Vanzin <vanzin@cloudera.com>
Closes#6022 from vanzin/SPARK-7485 and squashes the following commits:
22c1e04 [Marcelo Vanzin] Remove unneeded code.
4893622 [Marcelo Vanzin] [SPARK-7485] [build] Remove pyspark files from assembly.
JIRA issue [here](https://issues.apache.org/jira/browse/SPARK-5893).
One thing to make clear, the `buckets` parameter, which is an array of `Double`, performs as split points. Say,
```scala
buckets = Array(-0.5, 0.0, 0.5)
```
splits the real number into 4 ranges, (-inf, -0.5], (-0.5, 0.0], (0.0, 0.5], (0.5, +inf), which is encoded as 0, 1, 2, 3.
Author: Xusen Yin <yinxusen@gmail.com>
Author: Joseph K. Bradley <joseph@databricks.com>
Closes#5980 from yinxusen/SPARK-5893 and squashes the following commits:
dc8c843 [Xusen Yin] Merge pull request #4 from jkbradley/yinxusen-SPARK-5893
1ca973a [Joseph K. Bradley] one more bucketizer test
34f124a [Joseph K. Bradley] Removed lowerInclusive, upperInclusive params from Bucketizer, and used splits instead.
eacfcfa [Xusen Yin] change ML attribute from splits into buckets
c3cc770 [Xusen Yin] add more unit test for binary search
3a16cc2 [Xusen Yin] refine comments and names
ac77859 [Xusen Yin] fix style error
fb30d79 [Xusen Yin] fix and test binary search
2466322 [Xusen Yin] refactor Bucketizer
11fb00a [Xusen Yin] change it into an Estimator
998bc87 [Xusen Yin] check buckets
4024cf1 [Xusen Yin] add test suite
5fe190e [Xusen Yin] add bucketizer
Author: Yanbo Liang <ybliang8@gmail.com>
Closes#6044 from yanboliang/spark-6092 and squashes the following commits:
726a9b1 [Yanbo Liang] add newRankingMetrics
33f649c [Yanbo Liang] Add RankingMetrics in PySpark/MLlib
I implement a simple PCA wrapper for easy transform of vectors by PCA for example LabeledPoint or another complicated structure.
Example of usage:
```
import org.apache.spark.mllib.regression.LinearRegressionWithSGD
import org.apache.spark.mllib.regression.LabeledPoint
import org.apache.spark.mllib.linalg.Vectors
import org.apache.spark.mllib.feature.PCA
val data = sc.textFile("data/mllib/ridge-data/lpsa.data").map { line =>
val parts = line.split(',')
LabeledPoint(parts(0).toDouble, Vectors.dense(parts(1).split(' ').map(_.toDouble)))
}.cache()
val splits = data.randomSplit(Array(0.6, 0.4), seed = 11L)
val training = splits(0).cache()
val test = splits(1)
val pca = PCA.create(training.first().features.size/2, data.map(_.features))
val training_pca = training.map(p => p.copy(features = pca.transform(p.features)))
val test_pca = test.map(p => p.copy(features = pca.transform(p.features)))
val numIterations = 100
val model = LinearRegressionWithSGD.train(training, numIterations)
val model_pca = LinearRegressionWithSGD.train(training_pca, numIterations)
val valuesAndPreds = test.map { point =>
val score = model.predict(point.features)
(score, point.label)
}
val valuesAndPreds_pca = test_pca.map { point =>
val score = model_pca.predict(point.features)
(score, point.label)
}
val MSE = valuesAndPreds.map{case(v, p) => math.pow((v - p), 2)}.mean()
val MSE_pca = valuesAndPreds_pca.map{case(v, p) => math.pow((v - p), 2)}.mean()
println("Mean Squared Error = " + MSE)
println("PCA Mean Squared Error = " + MSE_pca)
```
Author: Kirill A. Korinskiy <catap@catap.ru>
Author: Joseph K. Bradley <joseph@databricks.com>
Closes#4304 from catap/pca and squashes the following commits:
501bcd9 [Joseph K. Bradley] Small updates: removed k from Java-friendly PCA fit(). In PCASuite, converted results to set for comparison. Added an error message for bad k in PCA.
9dcc02b [Kirill A. Korinskiy] [SPARK-5521] fix scala style
1892a06 [Kirill A. Korinskiy] [SPARK-5521] PCA wrapper for easy transform vectors
In SPARK-7429 and PR https://github.com/apache/spark/pull/5960, I added the varargs annotation to Params.setDefault which takes a variable number of ParamPairs. It worked locally and on Jenkins for me.
However, mengxr reported issues compiling on his machine. So I'm reverting the change introduced in https://github.com/apache/spark/pull/5960 by removing varargs.
Author: Joseph K. Bradley <joseph@databricks.com>
Closes#6021 from jkbradley/revert-varargs and squashes the following commits:
098ed39 [Joseph K. Bradley] removed varargs annotation from Params.setDefaults taking multiple ParamPairs
1) Handle scaling and addBias internally.
2) L1/L2 elasticnet using OWLQN optimizer.
Author: DB Tsai <dbt@netflix.com>
Closes#5967 from dbtsai/lor and squashes the following commits:
fa029bb [DB Tsai] made the bound smaller
0806002 [DB Tsai] better initial intercept and more test
5c31824 [DB Tsai] fix import
c387e25 [DB Tsai] Merge branch 'master' into lor
c84e931 [DB Tsai] Made MultiClassSummarizer private
f98e711 [DB Tsai] address feedback
a784321 [DB Tsai] fix style
8ec65d2 [DB Tsai] remove new line
f3f8c88 [DB Tsai] add more tests and they match R which is good. fix a bug
34705bc [DB Tsai] first commit
Adds Python Api for `ALS` under `ml.recommendation` in PySpark. Also adds seed as a settable parameter in the Scala Implementation of ALS.
Author: Burak Yavuz <brkyvz@gmail.com>
Closes#6015 from brkyvz/ml-rec and squashes the following commits:
be6e931 [Burak Yavuz] addressed comments
eaed879 [Burak Yavuz] readd numFeatures
0bd66b1 [Burak Yavuz] fixed seed
7f6d964 [Burak Yavuz] merged master
52e2bda [Burak Yavuz] added ALS
Add a Python API for mllib.feature.ChiSqSelector
https://issues.apache.org/jira/browse/SPARK-5913
Author: Yanbo Liang <ybliang8@gmail.com>
Closes#5939 from yanboliang/spark-5913 and squashes the following commits:
cdaac99 [Yanbo Liang] Python API for ChiSqSelector
Implemented python wrappers for Scala functions that don't exist in `ml.features`
Author: Burak Yavuz <brkyvz@gmail.com>
Closes#5991 from brkyvz/ml-feat-PR and squashes the following commits:
adcca55 [Burak Yavuz] add regex tokenizer to __all__
b91cb44 [Burak Yavuz] addressed comments
bd39fd2 [Burak Yavuz] remove addition
b82bd7c [Burak Yavuz] Parity in PySpark for ml.features
the toArray function of the BoundedPriorityQueue does not necessarily preserve order. Add a counter-example as the test, which would fail the original impl.
Author: Shuo Xiang <shuoxiangpub@gmail.com>
Closes#5990 from coderxiang/topbykey-test and squashes the following commits:
98804c9 [Shuo Xiang] fix bug in topBykey and update test
The compression is based on storage. brkyvz
Author: Xiangrui Meng <meng@databricks.com>
Closes#5985 from mengxr/SPARK-6948 and squashes the following commits:
df56a00 [Xiangrui Meng] update python tests
6d90d45 [Xiangrui Meng] compress vectors in VectorAssembler
https://issues.apache.org/jira/browse/SPARK-6093
Author: Yanbo Liang <ybliang8@gmail.com>
Closes#5941 from yanboliang/spark-6093 and squashes the following commits:
6934af3 [Yanbo Liang] change to @property
aac3bc5 [Yanbo Liang] Add RegressionMetrics in PySpark/MLlib
The wrapper required the implementation of the `ArrayParam`, because `Array[T]` is hard to obtain from Python. `ArrayParam` has an extra function called `wCast` which is an internal function to obtain `Array[T]` from `Seq[T]`
Author: Burak Yavuz <brkyvz@gmail.com>
Author: Xiangrui Meng <meng@databricks.com>
Closes#5930 from brkyvz/ml-feat and squashes the following commits:
73e745f [Burak Yavuz] Merge pull request #3 from mengxr/SPARK-7388
c221db9 [Xiangrui Meng] overload StringArrayParam.w
c81072d [Burak Yavuz] addressed comments
99c2ebf [Burak Yavuz] add to python_shared_params
39ecb07 [Burak Yavuz] fix scalastyle
7f7ea2a [Burak Yavuz] [SPARK-7388][SPARK-7383] wrapper for VectorAssembler in Python
Params.setDefault taking a set of ParamPairs should be annotated with varargs. I thought it would not work before, but it apparently does.
CrossValidator.transform should call transformSchema since the underlying Model might be a PipelineModel
CC: mengxr
Author: Joseph K. Bradley <joseph@databricks.com>
Closes#5960 from jkbradley/params-cleanups and squashes the following commits:
118b158 [Joseph K. Bradley] Params.setDefault taking a set of ParamPairs should be annotated with varargs. I thought it would not work before, but it apparently does. CrossValidator.transform should call transformSchema since the underlying Model might be a PipelineModel
Small changes, primarily to allow us more flexibility in the future:
* Rename "tau_0" to "tau0"
* Mark LDAOptimizer trait sealed and DeveloperApi.
* Mark LDAOptimizer subclasses as final.
* Mark setOptimizer (the one taking an LDAOptimizer) and getOptimizer as DeveloperApi since we may need to change them in the future
CC: hhbyyh
Author: Joseph K. Bradley <joseph@databricks.com>
Closes#5956 from jkbradley/onlinelda-cleanups and squashes the following commits:
f4be508 [Joseph K. Bradley] added newline
f4003e4 [Joseph K. Bradley] Changes: * Rename "tau_0" to "tau0" * Mark LDAOptimizer trait sealed and DeveloperApi. * Mark LDAOptimizer subclasses as final. * Mark setOptimizer (the one taking an LDAOptimizer) and getOptimizer as DeveloperApi since we may need to change them in the future
Changes:
* Update protected prediction methods, following design doc. **<--most interesting change**
* Changed abstract classes for Estimator and Model to be public. Added DeveloperApi tag. (I kept the traits for Estimator/Model Params private.)
* Changed ProbabilisticClassificationModel method names to use probability instead of probabilities.
CC: mengxr shivaram etrain
Author: Joseph K. Bradley <joseph@databricks.com>
Closes#5913 from jkbradley/public-dev-api and squashes the following commits:
e9aa0ea [Joseph K. Bradley] moved findMax to DenseVector and renamed to argmax. fixed bug for vector of length 0
15b9957 [Joseph K. Bradley] renamed probabilities to probability in method names
5cda84d [Joseph K. Bradley] regenerated sharedParams
7d1877a [Joseph K. Bradley] Made spark.ml prediction abstractions public. Organized their prediction methods for efficient computation of multiple output columns.
Since CrossValidator is a meta algorithm, we copy the implementation in Python. jkbradley
Author: Xiangrui Meng <meng@databricks.com>
Closes#5926 from mengxr/SPARK-6940 and squashes the following commits:
6af181f [Xiangrui Meng] add TODOs
8285134 [Xiangrui Meng] update doc
060f7c3 [Xiangrui Meng] update doctest
acac727 [Xiangrui Meng] add keyword args
cdddecd [Xiangrui Meng] add CrossValidator in Python
This patch adds a one hot encoder for categorical features. Planning to add documentation and another test after getting feedback on the approach.
A couple choices made here:
* There's an `includeFirst` option which, if false, creates numCategories - 1 columns and, if true, creates numCategories columns. The default is true, which is the behavior in scikit-learn.
* The user is expected to pass a `Seq` of category names when instantiating a `OneHotEncoder`. These can be easily gotten from a `StringIndexer`. The names are used for the output column names, which take the form colName_categoryName.
Author: Sandy Ryza <sandy@cloudera.com>
Closes#5500 from sryza/sandy-spark-5888 and squashes the following commits:
f383250 [Sandy Ryza] Infer label names automatically
6e257b9 [Sandy Ryza] Review comments
7c539cf [Sandy Ryza] Vector transformers
1c182dd [Sandy Ryza] SPARK-5888. [MLLIB]. Add OneHotEncoder as a Transformer
Require non empty input rdd such that we can take the first labeledpoint and get the feature size
Author: Alain <aihe@usc.edu>
Author: aihe@usc.edu <aihe@usc.edu>
Closes#5810 from AiHe/decisiontree-issue and squashes the following commits:
3b1d08a [aihe@usc.edu] [MLLIB][tree] merge the assertion into the evaluation of numFeatures
cf2e567 [Alain] [MLLIB][tree] Use a rdd api to verify size of input rdd > 0 when building meta data
b448f47 [Alain] [MLLIB][tree] Verify size of input rdd > 0 when building meta data
The following items are added to Python kmeans:
kmeans - setEpsilon, setInitializationSteps
KMeansModel - computeCost, k
Author: Hrishikesh Subramonian <hrishikesh.subramonian@flytxt.com>
Closes#5647 from FlytxtRnD/newPyKmeansAPI and squashes the following commits:
b9e451b [Hrishikesh Subramonian] set seed to fixed value in doc test
5fd3ced [Hrishikesh Subramonian] doc test corrections
20b3c68 [Hrishikesh Subramonian] python 3 fixes
4d4e695 [Hrishikesh Subramonian] added arguments in python tests
21eb84c [Hrishikesh Subramonian] Python Kmeans - setEpsilon, setInitializationSteps, k and computeCost added.
Utilities for pickling and unpickling SparseMatrices using SerDe
Author: MechCoder <manojkumarsivaraj334@gmail.com>
Closes#5775 from MechCoder/spark-7202 and squashes the following commits:
7e689dc [MechCoder] [SPARK-7202] Add SparseMatrixPickler to SerDe
This PR added `copy(extra: ParamMap): Params` to `Params`, which makes a copy of the current instance with a randomly generated uid and some extra param values. With this change, we only need to implement `fit` and `transform` without extra param values given the default implementation of `fit(dataset, extra)`:
~~~scala
def fit(dataset: DataFrame, extra: ParamMap): Model = {
copy(extra).fit(dataset)
}
~~~
Inside `fit` and `transform`, since only the embedded values are used, I added `$` as an alias for `getOrDefault` to make the code easier to read. For example, in `LinearRegression.fit` we have:
~~~scala
val effectiveRegParam = $(regParam) / yStd
val effectiveL1RegParam = $(elasticNetParam) * effectiveRegParam
val effectiveL2RegParam = (1.0 - $(elasticNetParam)) * effectiveRegParam
~~~
Meta-algorithm like `Pipeline` implements its own `copy(extra)`. So the fitted pipeline model stored all copied stages (no matter whether it is a transformer or a model).
Other changes:
* `Params$.inheritValues` is moved to `Params!.copyValues` and returns the target instance.
* `fittingParamMap` was removed because the `parent` carries this information.
* `validate` was renamed to `validateParams` to be more precise.
TODOs:
* [x] add tests for newly added methods
* [ ] update documentation
jkbradley dbtsai
Author: Xiangrui Meng <meng@databricks.com>
Closes#5820 from mengxr/SPARK-5956 and squashes the following commits:
7bef88d [Xiangrui Meng] address comments
05229c3 [Xiangrui Meng] assert -> assertEquals
b2927b1 [Xiangrui Meng] organize imports
f14456b [Xiangrui Meng] Merge remote-tracking branch 'apache/master' into SPARK-5956
93e7924 [Xiangrui Meng] add tests for hasParam & copy
463ecae [Xiangrui Meng] merge master
2b954c3 [Xiangrui Meng] update Binarizer
465dd12 [Xiangrui Meng] Merge remote-tracking branch 'apache/master' into SPARK-5956
282a1a8 [Xiangrui Meng] fix test
819dd2d [Xiangrui Meng] merge master
b642872 [Xiangrui Meng] example code runs
5a67779 [Xiangrui Meng] examples compile
c76b4d1 [Xiangrui Meng] fix all unit tests
0f4fd64 [Xiangrui Meng] fix some tests
9286a22 [Xiangrui Meng] copyValues to trained models
53e0973 [Xiangrui Meng] move inheritValues to Params and rename it to copyValues
9ee004e [Xiangrui Meng] merge copy and copyWith; rename validate to validateParams
d882afc [Xiangrui Meng] test compile
f082a31 [Xiangrui Meng] make Params copyable and simply handling of extra params in all spark.ml components
JIRA: https://issues.apache.org/jira/browse/SPARK-5563
The PR contains the implementation for [Online LDA] (https://www.cs.princeton.edu/~blei/papers/HoffmanBleiBach2010b.pdf) based on the research of Matt Hoffman and David M. Blei, which provides an efficient option for LDA users. Major advantages for the algorithm are the stream compatibility and economic time/memory consumption due to the corpus split. For more details, please refer to the jira.
Online LDA can act as a fast option for LDA, and will be especially helpful for the users who needs a quick result or with large corpus.
Correctness test.
I have tested current PR with https://github.com/Blei-Lab/onlineldavb and the results are identical. I've uploaded the result and code to https://github.com/hhbyyh/LDACrossValidation.
Author: Yuhao Yang <hhbyyh@gmail.com>
Author: Joseph K. Bradley <joseph@databricks.com>
Closes#4419 from hhbyyh/ldaonline and squashes the following commits:
1045eec [Yuhao Yang] Merge pull request #2 from jkbradley/hhbyyh-ldaonline2
cf376ff [Joseph K. Bradley] For private vars needed for testing, I made them private and added accessors. Java doesn’t understand package-private tags, so this minimizes the issues Java users might encounter.
6149ca6 [Yuhao Yang] fix for setOptimizer
cf0007d [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into ldaonline
54cf8da [Yuhao Yang] some style change
68c2318 [Yuhao Yang] add a java ut
4041723 [Yuhao Yang] add ut
138bfed [Yuhao Yang] Merge pull request #1 from jkbradley/hhbyyh-ldaonline-update
9e910d9 [Joseph K. Bradley] small fix
61d60df [Joseph K. Bradley] Minor cleanups: * Update *Concentration parameter documentation * EM Optimizer: createVertices() does not need to be a function * OnlineLDAOptimizer: typos in doc * Clean up the core code for online LDA (Scala style)
a996a82 [Yuhao Yang] respond to comments
b1178cf [Yuhao Yang] fit into the optimizer framework
dbe3cff [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into ldaonline
15be071 [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into ldaonline
b29193b [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into ldaonline
d19ef55 [Yuhao Yang] change OnlineLDA to class
97b9e1a [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into ldaonline
e7bf3b0 [Yuhao Yang] move to seperate file
f367cc9 [Yuhao Yang] change to optimization
8cb16a6 [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into ldaonline
62405cc [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into ldaonline
02d0373 [Yuhao Yang] fix style in comment
f6d47ca [Yuhao Yang] Merge branch 'ldaonline' of https://github.com/hhbyyh/spark into ldaonline
d86cdec [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into ldaonline
a570c9a [Yuhao Yang] use sample to pick up batch
4a3f27e [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into ldaonline
e271eb1 [Yuhao Yang] remove non ascii
581c623 [Yuhao Yang] seperate API and adjust batch split
37af91a [Yuhao Yang] iMerge remote-tracking branch 'upstream/master' into ldaonline
20328d1 [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into ldaonline i
aa365d1 [Yuhao Yang] merge upstream master
3a06526 [Yuhao Yang] merge with new example
0dd3947 [Yuhao Yang] kMerge remote-tracking branch 'upstream/master' into ldaonline
0d0f3ee [Yuhao Yang] replace random split with sliding
fa408a8 [Yuhao Yang] ssMerge remote-tracking branch 'upstream/master' into ldaonline
45884ab [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into ldaonline s
f41c5ca [Yuhao Yang] style fix
26dca1b [Yuhao Yang] style fix and make class private
043e786 [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into ldaonline s Conflicts: mllib/src/main/scala/org/apache/spark/mllib/clustering/LDA.scala
d640d9c [Yuhao Yang] online lda initial checkin
This is based on #3098 from debasish83.
1. BLAS' GEMM is used to compute inner products.
2. Reverted changes to MovieLensALS. SPARK-4231 should be addressed in a separate PR.
3. ~~Fixed a bug in topByKey~~
Closes#3098
debasish83 coderxiang
Author: Debasish Das <debasish.das@one.verizon.com>
Author: Xiangrui Meng <meng@databricks.com>
Closes#5829 from mengxr/SPARK-3066 and squashes the following commits:
22e6a87 [Xiangrui Meng] topByKey was correct. update its usage
389b381 [Xiangrui Meng] fix indentation
49953de [Xiangrui Meng] Merge remote-tracking branch 'apache/master' into SPARK-3066
cb9799a [Xiangrui Meng] revert MovieLensALS
f864f5e [Xiangrui Meng] update test and fix a bug in topByKey
c5e0181 [Xiangrui Meng] use GEMM and topByKey
3a0c4eb [Debasish Das] updated with spark master
98fa424 [Debasish Das] updated with master
ee99571 [Debasish Das] addressed initial review comments;merged with master;added tests for batch predict APIs in matrix factorization
3f97c49 [Debasish Das] fixed spark coding style for imports
7163a5c [Debasish Das] Added API for batch user and product recommendation; MAP calculation for product recommendation per user using randomized split
d144f57 [Debasish Das] recommendAll API to MatrixFactorizationModel, uses topK finding using BoundedPriorityQueue similar to RDD.top
f38a1b5 [Debasish Das] use sampleByKey for per user sampling
10cbb37 [Debasish Das] provide ratio for topN product validation; generate MAP and prec@k metric for movielens dataset
9fa063e [Debasish Das] import scala.math.round
4bbae0f [Debasish Das] comments fixed as per scalastyle
cd3ab31 [Debasish Das] merged with AbstractParams serialization bug
9b3951f [Debasish Das] validate user/product on MovieLens dataset through user input and compute map measure along with rmse
Author: DB Tsai <dbt@netflix.com>
Closes#5809 from dbtsai/format and squashes the following commits:
6904eed [DB Tsai] triger jenkins
9146e19 [DB Tsai] initial commit
See PDF attached to the JIRA issue 1406.
The contribution is my original work and I license the work to the project under the project's open source license.
Author: Vincenzo Selvaggio <vselvaggio@hotmail.it>
Author: Xiangrui Meng <meng@databricks.com>
Author: selvinsource <vselvaggio@hotmail.it>
Closes#3062 from selvinsource/mllib_pmml_model_export_SPARK-1406 and squashes the following commits:
852aac6 [Vincenzo Selvaggio] [SPARK-1406] Update JPMML version to 1.1.15 in LICENSE file
085cf42 [Vincenzo Selvaggio] [SPARK-1406] Added Double Min and Max Fixed scala style
30165c4 [Vincenzo Selvaggio] [SPARK-1406] Fixed extreme cases for logit
7a5e0ec [Vincenzo Selvaggio] [SPARK-1406] Binary classification for SVM and Logistic Regression
cfcb596 [Vincenzo Selvaggio] [SPARK-1406] Throw IllegalArgumentException when exporting a multinomial logistic regression
25dce33 [Vincenzo Selvaggio] [SPARK-1406] Update code to latest pmml model
dea98ca [Vincenzo Selvaggio] [SPARK-1406] Exclude transitive dependency for pmml model
66b7c12 [Vincenzo Selvaggio] [SPARK-1406] Updated pmml model lib to 1.1.15, latest Java 6 compatible
a0a55f7 [Vincenzo Selvaggio] Merge pull request #2 from mengxr/SPARK-1406
3c22f79 [Xiangrui Meng] more code style
e2313df [Vincenzo Selvaggio] Merge pull request #1 from mengxr/SPARK-1406
472d757 [Xiangrui Meng] fix code style
1676e15 [Vincenzo Selvaggio] fixed scala issue
e2ffae8 [Vincenzo Selvaggio] fixed scala style
b8823b0 [Vincenzo Selvaggio] Merge remote-tracking branch 'upstream/master' into mllib_pmml_model_export_SPARK-1406
b25bbf7 [Vincenzo Selvaggio] [SPARK-1406] Added export of pmml to distributed file system using the spark context
7a949d0 [Vincenzo Selvaggio] [SPARK-1406] Fixed scala style
f46c75c [Vincenzo Selvaggio] [SPARK-1406] Added PMMLExportable to supported models
7b33b4e [Vincenzo Selvaggio] [SPARK-1406] Added a PMMLExportable interface Restructured code in a new package mllib.pmml Supported models implements the new PMMLExportable interface: LogisticRegression, SVM, KMeansModel, LinearRegression, RidgeRegression, Lasso
d559ec5 [Vincenzo Selvaggio] Merge remote-tracking branch 'upstream/master' into mllib_pmml_model_export_SPARK-1406
8fe12bb [Vincenzo Selvaggio] [SPARK-1406] Adjusted logistic regression export description and target categories
03bc3a5 [Vincenzo Selvaggio] added logistic regression
da2ec11 [Vincenzo Selvaggio] [SPARK-1406] added linear SVM PMML export
82f2131 [Vincenzo Selvaggio] Merge remote-tracking branch 'upstream/master' into mllib_pmml_model_export_SPARK-1406
19adf29 [Vincenzo Selvaggio] [SPARK-1406] Fixed scala style
1faf985 [Vincenzo Selvaggio] [SPARK-1406] Added target field to the regression model for completeness Adjusted unit test to deal with this change
3ae8ae5 [Vincenzo Selvaggio] [SPARK-1406] Adjusted imported order according to the guidelines
c67ce81 [Vincenzo Selvaggio] Merge remote-tracking branch 'upstream/master' into mllib_pmml_model_export_SPARK-1406
78515ec [Vincenzo Selvaggio] [SPARK-1406] added pmml export for LinearRegressionModel, RidgeRegressionModel and LassoModel
e29dfb9 [Vincenzo Selvaggio] removed version, by default is set to 4.2 (latest from jpmml) removed copyright
ae8b993 [Vincenzo Selvaggio] updated some commented tests to use the new ModelExporter object reordered the imports
df8a89e [Vincenzo Selvaggio] added pmml version to pmml model changed the copyright to spark
a1b4dc3 [Vincenzo Selvaggio] updated imports
834ca44 [Vincenzo Selvaggio] reordered the import accordingly to the guidelines
349a76b [Vincenzo Selvaggio] new helper object to serialize the models to pmml format
c3ef9b8 [Vincenzo Selvaggio] set it to private
6357b98 [Vincenzo Selvaggio] set it to private
e1eb251 [Vincenzo Selvaggio] removed serialization part, this will be part of the ModelExporter helper object
aba5ee1 [Vincenzo Selvaggio] fixed cluster export
cd6c07c [Vincenzo Selvaggio] fixed scala style to run tests
f75b988 [Vincenzo Selvaggio] Merge remote-tracking branch 'origin/master' into mllib_pmml_model_export_SPARK-1406
07a29bf [selvinsource] Update LICENSE
8841439 [Vincenzo Selvaggio] adjust scala style in order to compile
1433b11 [Vincenzo Selvaggio] complete suite tests
8e71b8d [Vincenzo Selvaggio] kmeans pmml export implementation
9bc494f [Vincenzo Selvaggio] added scala suite tests added saveLocalFile to ModelExport trait
226e184 [Vincenzo Selvaggio] added javadoc and export model type in case there is a need to support other types of export (not just PMML)
a0e3679 [Vincenzo Selvaggio] export and pmml export traits kmeans test implementation
Author: DB Tsai <dbt@netflix.com>
Closes#5794 from dbtsai/clean and squashes the following commits:
ad639dd [DB Tsai] Indentation
834d527 [DB Tsai] Some code clean up.
Main change: Added isValid field to Param. Modified all usages to use isValid when relevant. Added helper methods in ParamValidate.
Also overrode Params.validate() in:
* CrossValidator + model
* Pipeline + model
I made a few updates for the elastic net patch:
* I changed "tol" to "convergenceTol"
* I added some documentation
This PR is Scala + Java only. Python will be in a follow-up PR.
CC: mengxr
Author: Joseph K. Bradley <joseph@databricks.com>
Closes#5740 from jkbradley/enforce-validate and squashes the following commits:
ad9c6c1 [Joseph K. Bradley] re-generated sharedParams after merging with current master
76415e8 [Joseph K. Bradley] reverted convergenceTol to tol
af62f4b [Joseph K. Bradley] Removed changes to SparkBuild, python linalg. Fixed test failures. Renamed ParamValidate to ParamValidators. Removed explicit type from ParamValidators calls where possible.
bb2665a [Joseph K. Bradley] merged with elastic net pr
ecda302 [Joseph K. Bradley] fix rat tests, plus add a little doc
6895dfc [Joseph K. Bradley] small cleanups
069ac6d [Joseph K. Bradley] many cleanups
928fb84 [Joseph K. Bradley] Maybe done
a910ac7 [Joseph K. Bradley] still workin
6d60e2e [Joseph K. Bradley] Still workin
b987319 [Joseph K. Bradley] Partly done with adding checks, but blocking on adding checking functionality to Param
dbc9fb2 [Joseph K. Bradley] merged with master. enforcing Params.validate
Changed VectorIndexer so it does not carry non-ML metadata from the input to the output column. Removed ml.util.TestingUtils since VectorIndexer was the only use.
CC: mengxr
Author: Joseph K. Bradley <joseph@databricks.com>
Closes#5789 from jkbradley/vector-indexer-metadata and squashes the following commits:
b28e159 [Joseph K. Bradley] Changed VectorIndexer so it does not carry non-ML metadata from the input to the output column. Removed ml.util.TestingUtils since VectorIndexer was the only use.
See JIRA issue [here](https://issues.apache.org/jira/browse/SPARK-6529).
There are some notes:
1. I add `learningRate` in sharedParams since it is a common parameter for ML algorithms.
2. We will not support transform of finding synonyms from a `Vector`, which will support in further JIRA issues.
3. Word2Vec is different with other ML models that its training set and transformed set are different. Its training set is an `RDD[Iterable[String]]` which represents documents, but the transformed set we want is an `RDD[String]` that represents unique words. So you have to switch your `inputCol` in these two stages.
Author: Xusen Yin <yinxusen@gmail.com>
Closes#5596 from yinxusen/SPARK-6529 and squashes the following commits:
ee2b37a [Xusen Yin] merge with former HEAD
4945462 [Xusen Yin] merge with #5626
3bc2cbd [Xusen Yin] change foldLeft to for loop and use blas
5dd4ee7 [Xusen Yin] fix scala style
743e0d5 [Xusen Yin] fix comments and code style
04c48e9 [Xusen Yin] ensure the functionality
a190f2c [Xusen Yin] fix code style and refine the transform function of word2vec
02848fa [Xusen Yin] refine comments
34a55c0 [Xusen Yin] fix errors
109d124 [Xusen Yin] add test suite and pass it
04dde06 [Xusen Yin] add shared params
c594095 [Xusen Yin] add word2vec transformer
23d77fa [Xusen Yin] merge with #5626
e8cfaf7 [Xusen Yin] fix conflict with master
66e7bd3 [Xusen Yin] change foldLeft to for loop and use blas
566ec20 [Xusen Yin] fix scala style
b54399f [Xusen Yin] fix comments and code style
1211e86 [Xusen Yin] ensure the functionality
6b97ec8 [Xusen Yin] fix code style and refine the transform function of word2vec
7cde18f [Xusen Yin] rm sharedParams
618abd0 [Xusen Yin] refine comments
e29680a [Xusen Yin] fix errors
fe3afe9 [Xusen Yin] add test suite and pass it
02767fb [Xusen Yin] add shared params
6a514f1 [Xusen Yin] add word2vec transformer
Added detailed mathematical derivation of how scaling and LeastSquaresAggregator work. Refactored the code so the model is compressed based on the storage. We may try compression based on the prediction time.
Also, I found that diffSum will be always zero mathematically, so no corrections are required.
Author: DB Tsai <dbt@netflix.com>
Closes#5767 from dbtsai/lir-doc and squashes the following commits:
5e346c9 [DB Tsai] refactoring
fc9f582 [DB Tsai] doc
58456d8 [DB Tsai] address feedback
69757b8 [DB Tsai] actually diffSum is mathematically zero! No correction is needed.
5929e49 [DB Tsai] typo
63f7d1e [DB Tsai] Added compression to the model based on storage
203a295 [DB Tsai] Add more documentation to LinearRegression in new ML framework.
mengxr
Author: Xusen Yin <yinxusen@gmail.com>
Closes#5769 from yinxusen/patch-1 and squashes the following commits:
43235f4 [Xusen Yin] Update PearsonCorrelation.scala
f7287ee [Xusen Yin] Fix a typo of "threshold"
Add `compressed` to `Vector` with some other methods: `numActives`, `numNonzeros`, `toSparse`, and `toDense`. jkbradley
Author: Xiangrui Meng <meng@databricks.com>
Closes#5756 from mengxr/SPARK-6756 and squashes the following commits:
8d4ecbd [Xiangrui Meng] address comment and add mima excludes
da54179 [Xiangrui Meng] add toSparse, toDense, numActives, numNonzeros, and compressed to Vector
Cast numeric types to String for indexing. Boolean type is not handled in this PR. jkbradley
Author: Xiangrui Meng <meng@databricks.com>
Closes#5753 from mengxr/SPARK-6965 and squashes the following commits:
2e34f3c [Xiangrui Meng] add actual type in the error message
ad938bf [Xiangrui Meng] StringIndexer handles numeric input.
It shouldn't live directly under `spark.ml`.
Author: Xiangrui Meng <meng@databricks.com>
Closes#5749 from mengxr/SPARK-7201 and squashes the following commits:
53847f9 [Xiangrui Meng] move Identifiable to ml.util
The Python SerDe calls `Object.hashCode`, which is very expensive for Vectors. It is not necessary to scan the whole vector, especially for large ones. In this PR, we only scan the first 16 nonzeros. srowen
Author: Xiangrui Meng <meng@databricks.com>
Closes#5697 from mengxr/SPARK-7140 and squashes the following commits:
2abc86d [Xiangrui Meng] typo
8fb7d74 [Xiangrui Meng] update impl
1ebad60 [Xiangrui Meng] only scan the first 16 nonzeros in Vector.hashCode
Author: DB Tsai <dbt@netflix.com>
Author: DB Tsai <dbtsai@alpinenow.com>
Closes#4259 from dbtsai/lir and squashes the following commits:
a81c201 [DB Tsai] add import org.apache.spark.util.Utils back
9fc48ed [DB Tsai] rebase
2178b63 [DB Tsai] add comments
9988ca8 [DB Tsai] addressed feedback and fixed a bug. TODO: documentation and build another synthetic dataset which can catch the bug fixed in this commit.
fcbaefe [DB Tsai] Refactoring
4eb078d [DB Tsai] first commit
jira: https://issues.apache.org/jira/browse/SPARK-7090
LDA was implemented with extensibility in mind. And with the development of OnlineLDA and Gibbs Sampling, we are collecting more detailed requirements from different algorithms.
As Joseph Bradley jkbradley proposed in https://github.com/apache/spark/pull/4807 and with some further discussion, we'd like to adjust the code structure a little to present the common interface and extension point clearly.
Basically class LDA would be a common entrance for LDA computing. And each LDA object will refer to a LDAOptimizer for the concrete algorithm implementation. Users can customize LDAOptimizer with specific parameters and assign it to LDA.
Concrete changes:
1. Add a trait `LDAOptimizer`, which defines the common iterface for concrete implementations. Each subClass is a wrapper for a specific LDA algorithm.
2. Move EMOptimizer to file LDAOptimizer and inherits from LDAOptimizer, rename to EMLDAOptimizer. (in case a more generic EMOptimizer comes in the future)
-adjust the constructor of EMOptimizer, since all the parameters should be passed in through initialState method. This can avoid unwanted confusion or overwrite.
-move the code from LDA.initalState to initalState of EMLDAOptimizer
3. Add property ldaOptimizer to LDA and its getter/setter, and EMLDAOptimizer is the default Optimizer.
4. Change the return type of LDA.run from DistributedLDAModel to LDAModel.
Further work:
add OnlineLDAOptimizer and other possible Optimizers once ready.
Author: Yuhao Yang <hhbyyh@gmail.com>
Closes#5661 from hhbyyh/ldaRefactor and squashes the following commits:
0e2e006 [Yuhao Yang] respond to review comments
08a45da [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into ldaRefactor
e756ce4 [Yuhao Yang] solve mima exception
d74fd8f [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into ldaRefactor
0bb8400 [Yuhao Yang] refactor LDA with Optimizer
ec2f857 [Yuhao Yang] protoptype for discussion
