[jira] [Commented] (SPARK-12372) Document limitations of MLlib local linear algebra
[
https://issues.apache.org/jira/browse/SPARK-12372?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15064476#comment-15064476
]
Christos Iraklis Tsatsoulis commented on SPARK-12372:
-
You are very welcome
> Document limitations of MLlib local linear algebra
> --
>
> Key: SPARK-12372
> URL: https://issues.apache.org/jira/browse/SPARK-12372
> Project: Spark
> Issue Type: Documentation
> Components: Documentation, MLlib
>Affects Versions: 1.5.2
>Reporter: Christos Iraklis Tsatsoulis
>
> This JIRA is now for documenting limitations of MLlib's local linear algebra
> types. Basically, we should make it clear in the user guide that they
> provide simple functionality but are not a full-fledged local linear library.
> We should also recommend libraries for users to use in the meantime:
> probably Breeze for Scala (and Java?) and numpy/scipy for Python.
> *Original JIRA title*: Unary operator "-" fails for MLlib vectors
> *Original JIRA text, as an example of the need for better docs*:
> Consider the following snippet in pyspark 1.5.2:
> {code:none}
> >>> from pyspark.mllib.linalg import Vectors
> >>> x = Vectors.dense([0.0, 1.0, 0.0, 7.0, 0.0])
> >>> x
> DenseVector([0.0, 1.0, 0.0, 7.0, 0.0])
> >>> -x
> Traceback (most recent call last):
> File "", line 1, in
> TypeError: func() takes exactly 2 arguments (1 given)
> >>> y = Vectors.dense([2.0, 0.0, 3.0, 4.0, 5.0])
> >>> y
> DenseVector([2.0, 0.0, 3.0, 4.0, 5.0])
> >>> x-y
> DenseVector([-2.0, 1.0, -3.0, 3.0, -5.0])
> >>> -y+x
> Traceback (most recent call last):
> File "", line 1, in
> TypeError: func() takes exactly 2 arguments (1 given)
> >>> -1*x
> DenseVector([-0.0, -1.0, -0.0, -7.0, -0.0])
> {code}
> Clearly, the unary operator {{-}} (minus) for vectors fails, giving errors
> for expressions like {{-x}} and {{-y+x}}, despite the fact that {{x-y}}
> behaves as expected.
> The last operation, {{-1*x}}, although mathematically "correct", includes
> minus signs for the zero entries, which again is normally not expected.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-12372) Unary operator "-" fails for MLlib vectors
[
https://issues.apache.org/jira/browse/SPARK-12372?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15060858#comment-15060858
]
Christos Iraklis Tsatsoulis commented on SPARK-12372:
-
If this is the case, then a warning/clarification in the documentation wouldn't
hurt - Spark users are not supposed to be aware of the internal "ongoing
discussions" between Spark developers (BTW, any relevant link would be very
welcome - I could not find any mention in MLlib & Breeze docs, neither in the
recent preprint papers on linalg & MLlib).
All in all, I suggest you re-open the issue with a different type (it's not a
bug, as you say), and the required resolution being a notification in the
relevant docs ("don't try this..., because...").
> Unary operator "-" fails for MLlib vectors
> --
>
> Key: SPARK-12372
> URL: https://issues.apache.org/jira/browse/SPARK-12372
> Project: Spark
> Issue Type: Bug
> Components: MLlib, PySpark
>Affects Versions: 1.5.2
>Reporter: Christos Iraklis Tsatsoulis
>
> Consider the following snippet in pyspark 1.5.2:
> {code:none}
> >>> from pyspark.mllib.linalg import Vectors
> >>> x = Vectors.dense([0.0, 1.0, 0.0, 7.0, 0.0])
> >>> x
> DenseVector([0.0, 1.0, 0.0, 7.0, 0.0])
> >>> -x
> Traceback (most recent call last):
> File "", line 1, in
> TypeError: func() takes exactly 2 arguments (1 given)
> >>> y = Vectors.dense([2.0, 0.0, 3.0, 4.0, 5.0])
> >>> y
> DenseVector([2.0, 0.0, 3.0, 4.0, 5.0])
> >>> x-y
> DenseVector([-2.0, 1.0, -3.0, 3.0, -5.0])
> >>> -y+x
> Traceback (most recent call last):
> File "", line 1, in
> TypeError: func() takes exactly 2 arguments (1 given)
> >>> -1*x
> DenseVector([-0.0, -1.0, -0.0, -7.0, -0.0])
> {code}
> Clearly, the unary operator {{-}} (minus) for vectors fails, giving errors
> for expressions like {{-x}} and {{-y+x}}, despite the fact that {{x-y}}
> behaves as expected.
> The last operation, {{-1*x}}, although mathematically "correct", includes
> minus signs for the zero entries, which again is normally not expected.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Created] (SPARK-12372) Unary operator "-" fails for MLlib vectors
Christos Iraklis Tsatsoulis created SPARK-12372:
---
Summary: Unary operator "-" fails for MLlib vectors
Key: SPARK-12372
URL: https://issues.apache.org/jira/browse/SPARK-12372
Project: Spark
Issue Type: Bug
Components: MLlib, PySpark
Affects Versions: 1.5.2
Reporter: Christos Iraklis Tsatsoulis
Consider the following snippet in pyspark 1.5.2:
{code:none}
>>> from pyspark.mllib.linalg import Vectors
>>> x = Vectors.dense([0.0, 1.0, 0.0, 7.0, 0.0])
>>> x
DenseVector([0.0, 1.0, 0.0, 7.0, 0.0])
>>> -x
Traceback (most recent call last):
File "", line 1, in
TypeError: func() takes exactly 2 arguments (1 given)
>>> y = Vectors.dense([2.0, 0.0, 3.0, 4.0, 5.0])
>>> y
DenseVector([2.0, 0.0, 3.0, 4.0, 5.0])
>>> x-y
DenseVector([-2.0, 1.0, -3.0, 3.0, -5.0])
>>> -y+x
Traceback (most recent call last):
File "", line 1, in
TypeError: func() takes exactly 2 arguments (1 given)
>>> -1*x
DenseVector([-0.0, -1.0, -0.0, -7.0, -0.0])
{code}
Clearly, the unary operator {{-}} (minus) for vectors fails, giving errors for
expressions like {{-x}} and {{-y+x}}, despite the fact that {{x-y}} behaves as
expected.
The last operation, {{-1*x}}, although mathematically "correct", includes minus
signs for the zero entries, which again is normally not expected.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Updated] (SPARK-11530) Return eigenvalues with PCA model
[ https://issues.apache.org/jira/browse/SPARK-11530?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Christos Iraklis Tsatsoulis updated SPARK-11530: Component/s: MLlib > Return eigenvalues with PCA model > - > > Key: SPARK-11530 > URL: https://issues.apache.org/jira/browse/SPARK-11530 > Project: Spark > Issue Type: Improvement > Components: ML, MLlib >Affects Versions: 1.5.1 >Reporter: Christos Iraklis Tsatsoulis > > For data scientists & statisticians, PCA is of little use if they cannot > estimate the _proportion of variance explained_ by selecting _k_ principal > components (see here for the math details: > https://inst.eecs.berkeley.edu/~ee127a/book/login/l_sym_pca.html , section > 'Explained variance'). To estimate this, one only needs the eigenvalues of > the covariance matrix. > Although the eigenvalues are currently computed during PCA model fitting, > they are not _returned_; hence, as it stands now, PCA in Spark ML is of > extremely limited practical use. > For details, see these SO questions > http://stackoverflow.com/questions/33428589/pyspark-and-pca-how-can-i-extract-the-eigenvectors-of-this-pca-how-can-i-calcu/ > (pyspark) > http://stackoverflow.com/questions/33559599/spark-pca-top-components (Scala) > and this blog post http://www.nodalpoint.com/pca-in-spark-1-5/ -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Comment Edited] (SPARK-11530) Return eigenvalues with PCA model
[ https://issues.apache.org/jira/browse/SPARK-11530?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=14996533#comment-14996533 ] Christos Iraklis Tsatsoulis edited comment on SPARK-11530 at 11/9/15 1:40 PM: -- I edited it to target both; there are `PCA.scala` scripts for both ML & MLLib, but since I am using it via PySpark, where it is available only via ML, I initially omitted MLlib was (Author: ctsats): I edited it to target both; there are PCA.scala scripts for both ML & MLLib, but since I am using it via PySpark, where it is available only via ML, I initially omitted MLlib > Return eigenvalues with PCA model > - > > Key: SPARK-11530 > URL: https://issues.apache.org/jira/browse/SPARK-11530 > Project: Spark > Issue Type: Improvement > Components: ML, MLlib >Affects Versions: 1.5.1 >Reporter: Christos Iraklis Tsatsoulis > > For data scientists & statisticians, PCA is of little use if they cannot > estimate the _proportion of variance explained_ by selecting _k_ principal > components (see here for the math details: > https://inst.eecs.berkeley.edu/~ee127a/book/login/l_sym_pca.html , section > 'Explained variance'). To estimate this, one only needs the eigenvalues of > the covariance matrix. > Although the eigenvalues are currently computed during PCA model fitting, > they are not _returned_; hence, as it stands now, PCA in Spark ML is of > extremely limited practical use. > For details, see these SO questions > http://stackoverflow.com/questions/33428589/pyspark-and-pca-how-can-i-extract-the-eigenvectors-of-this-pca-how-can-i-calcu/ > (pyspark) > http://stackoverflow.com/questions/33559599/spark-pca-top-components (Scala) > and this blog post http://www.nodalpoint.com/pca-in-spark-1-5/ -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Comment Edited] (SPARK-11530) Return eigenvalues with PCA model
[ https://issues.apache.org/jira/browse/SPARK-11530?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=14996533#comment-14996533 ] Christos Iraklis Tsatsoulis edited comment on SPARK-11530 at 11/9/15 1:45 PM: -- I edited it to target both; there are PCA.scala scripts for both ML & MLLib, but since I am using it via PySpark, where it is available only via ML, I initially omitted MLlib was (Author: ctsats): I edited it to target both; there are `PCA.scala` scripts for both ML & MLLib, but since I am using it via PySpark, where it is available only via ML, I initially omitted MLlib > Return eigenvalues with PCA model > - > > Key: SPARK-11530 > URL: https://issues.apache.org/jira/browse/SPARK-11530 > Project: Spark > Issue Type: Improvement > Components: ML, MLlib >Affects Versions: 1.5.1 >Reporter: Christos Iraklis Tsatsoulis > > For data scientists & statisticians, PCA is of little use if they cannot > estimate the _proportion of variance explained_ by selecting _k_ principal > components (see here for the math details: > https://inst.eecs.berkeley.edu/~ee127a/book/login/l_sym_pca.html , section > 'Explained variance'). To estimate this, one only needs the eigenvalues of > the covariance matrix. > Although the eigenvalues are currently computed during PCA model fitting, > they are not _returned_; hence, as it stands now, PCA in Spark ML is of > extremely limited practical use. > For details, see these SO questions > http://stackoverflow.com/questions/33428589/pyspark-and-pca-how-can-i-extract-the-eigenvectors-of-this-pca-how-can-i-calcu/ > (pyspark) > http://stackoverflow.com/questions/33559599/spark-pca-top-components (Scala) > and this blog post http://www.nodalpoint.com/pca-in-spark-1-5/ -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Comment Edited] (SPARK-11530) Return eigenvalues with PCA model
[ https://issues.apache.org/jira/browse/SPARK-11530?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=14996533#comment-14996533 ] Christos Iraklis Tsatsoulis edited comment on SPARK-11530 at 11/9/15 1:50 PM: -- I edited it to target both; there are PCA.scala scripts for both ML & MLlib, but since I am using it via PySpark, where it is available only via ML, I initially omitted MLlib. was (Author: ctsats): I edited it to target both; there are PCA.scala scripts for both ML & MLLib, but since I am using it via PySpark, where it is available only via ML, I initially omitted MLlib > Return eigenvalues with PCA model > - > > Key: SPARK-11530 > URL: https://issues.apache.org/jira/browse/SPARK-11530 > Project: Spark > Issue Type: Improvement > Components: ML, MLlib >Affects Versions: 1.5.1 >Reporter: Christos Iraklis Tsatsoulis > > For data scientists & statisticians, PCA is of little use if they cannot > estimate the _proportion of variance explained_ by selecting _k_ principal > components (see here for the math details: > https://inst.eecs.berkeley.edu/~ee127a/book/login/l_sym_pca.html , section > 'Explained variance'). To estimate this, one only needs the eigenvalues of > the covariance matrix. > Although the eigenvalues are currently computed during PCA model fitting, > they are not _returned_; hence, as it stands now, PCA in Spark ML is of > extremely limited practical use. > For details, see these SO questions > http://stackoverflow.com/questions/33428589/pyspark-and-pca-how-can-i-extract-the-eigenvectors-of-this-pca-how-can-i-calcu/ > (pyspark) > http://stackoverflow.com/questions/33559599/spark-pca-top-components (Scala) > and this blog post http://www.nodalpoint.com/pca-in-spark-1-5/ -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-11530) Return eigenvalues with PCA model
[ https://issues.apache.org/jira/browse/SPARK-11530?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=14996533#comment-14996533 ] Christos Iraklis Tsatsoulis commented on SPARK-11530: - I edited it to target both; there are ``PCA.scala`` scripts for both ML & MLLib, but since I am using it via PySpark, where it is available only via ML, I initially omitted MLlib > Return eigenvalues with PCA model > - > > Key: SPARK-11530 > URL: https://issues.apache.org/jira/browse/SPARK-11530 > Project: Spark > Issue Type: Improvement > Components: ML, MLlib >Affects Versions: 1.5.1 >Reporter: Christos Iraklis Tsatsoulis > > For data scientists & statisticians, PCA is of little use if they cannot > estimate the _proportion of variance explained_ by selecting _k_ principal > components (see here for the math details: > https://inst.eecs.berkeley.edu/~ee127a/book/login/l_sym_pca.html , section > 'Explained variance'). To estimate this, one only needs the eigenvalues of > the covariance matrix. > Although the eigenvalues are currently computed during PCA model fitting, > they are not _returned_; hence, as it stands now, PCA in Spark ML is of > extremely limited practical use. > For details, see these SO questions > http://stackoverflow.com/questions/33428589/pyspark-and-pca-how-can-i-extract-the-eigenvectors-of-this-pca-how-can-i-calcu/ > (pyspark) > http://stackoverflow.com/questions/33559599/spark-pca-top-components (Scala) > and this blog post http://www.nodalpoint.com/pca-in-spark-1-5/ -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Comment Edited] (SPARK-11530) Return eigenvalues with PCA model
[ https://issues.apache.org/jira/browse/SPARK-11530?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=14996533#comment-14996533 ] Christos Iraklis Tsatsoulis edited comment on SPARK-11530 at 11/9/15 1:37 PM: -- I edited it to target both; there are PCA.scala scripts for both ML & MLLib, but since I am using it via PySpark, where it is available only via ML, I initially omitted MLlib was (Author: ctsats): I edited it to target both; there are ``PCA.scala`` scripts for both ML & MLLib, but since I am using it via PySpark, where it is available only via ML, I initially omitted MLlib > Return eigenvalues with PCA model > - > > Key: SPARK-11530 > URL: https://issues.apache.org/jira/browse/SPARK-11530 > Project: Spark > Issue Type: Improvement > Components: ML, MLlib >Affects Versions: 1.5.1 >Reporter: Christos Iraklis Tsatsoulis > > For data scientists & statisticians, PCA is of little use if they cannot > estimate the _proportion of variance explained_ by selecting _k_ principal > components (see here for the math details: > https://inst.eecs.berkeley.edu/~ee127a/book/login/l_sym_pca.html , section > 'Explained variance'). To estimate this, one only needs the eigenvalues of > the covariance matrix. > Although the eigenvalues are currently computed during PCA model fitting, > they are not _returned_; hence, as it stands now, PCA in Spark ML is of > extremely limited practical use. > For details, see these SO questions > http://stackoverflow.com/questions/33428589/pyspark-and-pca-how-can-i-extract-the-eigenvectors-of-this-pca-how-can-i-calcu/ > (pyspark) > http://stackoverflow.com/questions/33559599/spark-pca-top-components (Scala) > and this blog post http://www.nodalpoint.com/pca-in-spark-1-5/ -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Updated] (SPARK-11530) Return eigenvalues with PCA model
[ https://issues.apache.org/jira/browse/SPARK-11530?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Christos Iraklis Tsatsoulis updated SPARK-11530: Description: For data scientists & statisticians, PCA is of little use if they cannot estimate the _proportion of variance explained_ by selecting _k_ principal components (see here for the math details: https://inst.eecs.berkeley.edu/~ee127a/book/login/l_sym_pca.html , section 'Explained variance'). To estimate this, one only needs the eigenvalues of the covariance matrix. Although the eigenvalues are currently computed during PCA model fitting, they are not _returned_; hence, as it stands now, PCA in Spark ML is of extremely limited practical use. For details, see these SO questions http://stackoverflow.com/questions/33428589/pyspark-and-pca-how-can-i-extract-the-eigenvectors-of-this-pca-how-can-i-calcu/ (pyspark) http://stackoverflow.com/questions/33559599/spark-pca-top-components (Scala) and this blog post http://www.nodalpoint.com/pca-in-spark-1-5/ was: For data scientists & statisticians, PCA is of little use if they cannot estimate the _proportion of variance explained_ by selecting _k_ principal components (see here for the math details: https://inst.eecs.berkeley.edu/~ee127a/book/login/l_sym_pca.html , section 'Explained variance'). To estimate this, one only needs the eigenvalues of the covariance matrix. Although the eigenvalues are currently computed during PCA model fitting, they are not _returned_; hence, as it stands now, PCA in Spark ML is of extremely limited practical use. See this SO question http://stackoverflow.com/questions/33428589/pyspark-and-pca-how-can-i-extract-the-eigenvectors-of-this-pca-how-can-i-calcu/ and this blog post http://www.nodalpoint.com/pca-in-spark-1-5/ for details. > Return eigenvalues with PCA model > - > > Key: SPARK-11530 > URL: https://issues.apache.org/jira/browse/SPARK-11530 > Project: Spark > Issue Type: Improvement > Components: ML >Affects Versions: 1.5.1 >Reporter: Christos Iraklis Tsatsoulis > > For data scientists & statisticians, PCA is of little use if they cannot > estimate the _proportion of variance explained_ by selecting _k_ principal > components (see here for the math details: > https://inst.eecs.berkeley.edu/~ee127a/book/login/l_sym_pca.html , section > 'Explained variance'). To estimate this, one only needs the eigenvalues of > the covariance matrix. > Although the eigenvalues are currently computed during PCA model fitting, > they are not _returned_; hence, as it stands now, PCA in Spark ML is of > extremely limited practical use. > For details, see these SO questions > http://stackoverflow.com/questions/33428589/pyspark-and-pca-how-can-i-extract-the-eigenvectors-of-this-pca-how-can-i-calcu/ > (pyspark) > http://stackoverflow.com/questions/33559599/spark-pca-top-components (Scala) > and this blog post http://www.nodalpoint.com/pca-in-spark-1-5/ -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-11530) Return eigenvalues with PCA model
[ https://issues.apache.org/jira/browse/SPARK-11530?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=14994136#comment-14994136 ] Christos Iraklis Tsatsoulis commented on SPARK-11530: - Thanks Sean. Unfortunately, I don't speak Scala - I'm actually a data scientist and not a developer. Thought that it was worthy raising the issue, even if I cannot resolve it myself. Hope that's OK... > Return eigenvalues with PCA model > - > > Key: SPARK-11530 > URL: https://issues.apache.org/jira/browse/SPARK-11530 > Project: Spark > Issue Type: Improvement > Components: ML >Affects Versions: 1.5.1 >Reporter: Christos Iraklis Tsatsoulis > > For data scientists & statisticians, PCA is of little use if they cannot > estimate the _proportion of variance explained_ by selecting _k_ principal > components (see here for the math details: > https://inst.eecs.berkeley.edu/~ee127a/book/login/l_sym_pca.html , section > 'Explained variance'). To estimate this, one only needs the eigenvalues of > the covariance matrix. > Although the eigenvalues are currently computed during PCA model fitting, > they are not _returned_; hence, as it stands now, PCA in Spark ML is of > extremely limited practical use. > For details, see these SO questions > http://stackoverflow.com/questions/33428589/pyspark-and-pca-how-can-i-extract-the-eigenvectors-of-this-pca-how-can-i-calcu/ > (pyspark) > http://stackoverflow.com/questions/33559599/spark-pca-top-components (Scala) > and this blog post http://www.nodalpoint.com/pca-in-spark-1-5/ -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Updated] (SPARK-11530) Return eigenvalues with PCA model
[ https://issues.apache.org/jira/browse/SPARK-11530?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Christos Iraklis Tsatsoulis updated SPARK-11530: Description: For data scientists & statisticians, PCA is of little use if they cannot estimate the _proportion of variance explained_ by selecting _k_ principal components (see here for the math details: https://inst.eecs.berkeley.edu/~ee127a/book/login/l_sym_pca.html , section 'Explained variance'). To estimate this, one only needs the eigenvalues of the covariance matrix. Although the eigenvalues are currently computed during PCA model fitting, they are not _returned_; hence, as it stands now, PCA in Spark ML is of extremely limited practical use. See this SO question http://stackoverflow.com/questions/33428589/pyspark-and-pca-how-can-i-extract-the-eigenvectors-of-this-pca-how-can-i-calcu/ and this blog post http://www.nodalpoint.com/pca-in-spark-1-5/ for details. was: For data scientists & statisticians, PCA is of little use if they cannot estimate the _proportion of variance explained_ by selecting _k_ principal components (see here for the math details: https://inst.eecs.berkeley.edu/~ee127a/book/login/l_sym_pca.html , section 'Explained variance'). To estimate this, one only needs the eigenvalues of the covariance matrix. Although the eigenvalues are currently computed during PCA model fitting, they are not _returned_; hence, as it stands now, PCA in Spark ML is of extremely limited practical use. See this SO question http://stackoverflow.com/questions/33428589/pyspark-and-pca-how-can-i-extract-the-eigenvectors-of-this-pca-how-can-i-calcu/) and this blog post http://www.nodalpoint.com/pca-in-spark-1-5/ for details. > Return eigenvalues with PCA model > - > > Key: SPARK-11530 > URL: https://issues.apache.org/jira/browse/SPARK-11530 > Project: Spark > Issue Type: Improvement > Components: ML >Affects Versions: 1.5.1 >Reporter: Christos Iraklis Tsatsoulis > > For data scientists & statisticians, PCA is of little use if they cannot > estimate the _proportion of variance explained_ by selecting _k_ principal > components (see here for the math details: > https://inst.eecs.berkeley.edu/~ee127a/book/login/l_sym_pca.html , section > 'Explained variance'). To estimate this, one only needs the eigenvalues of > the covariance matrix. > Although the eigenvalues are currently computed during PCA model fitting, > they are not _returned_; hence, as it stands now, PCA in Spark ML is of > extremely limited practical use. > See this SO question > http://stackoverflow.com/questions/33428589/pyspark-and-pca-how-can-i-extract-the-eigenvectors-of-this-pca-how-can-i-calcu/ > > and this blog post http://www.nodalpoint.com/pca-in-spark-1-5/ > for details. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
