[
https://issues.apache.org/jira/browse/SPARK-19078?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
zhengruifeng updated SPARK-19078:
---------------------------------
Description:
There some ML.algs (hashingTF,ChiSqSelector,IDF,StandardScaler,PCA,LDA) that
use oldModel in MLlib in {{transform}} method, this will cause extra conversion
of vectors and matrices.
This pr modify transform to direct use ml.linalg in
{{hashingTF,ChiSqSelector,IDF,StandardScaler,PCA}}.
I do not touch LDA because it is more complicated than others.
was:
As suggested in the source, avoid the vector conversion in
{{PCAModel.transform()}}
{{// TODO: Make the transformer natively in ml framework to avoid extra
conversion.}}
> ML.transform avoid extra vector conversion
> ------------------------------------------
>
> Key: SPARK-19078
> URL: https://issues.apache.org/jira/browse/SPARK-19078
> Project: Spark
> Issue Type: Improvement
> Components: ML
> Reporter: zhengruifeng
> Priority: Trivial
>
> There some ML.algs (hashingTF,ChiSqSelector,IDF,StandardScaler,PCA,LDA) that
> use oldModel in MLlib in {{transform}} method, this will cause extra
> conversion of vectors and matrices.
> This pr modify transform to direct use ml.linalg in
> {{hashingTF,ChiSqSelector,IDF,StandardScaler,PCA}}.
> I do not touch LDA because it is more complicated than others.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
