[jira] [Commented] (SPARK-5089) Vector conversion broken for non-float64 arrays

["Apache Spark (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SPARK-5089?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14265025#comment-14265025 ] 

Apache Spark commented on SPARK-5089:
-------------------------------------

User 'freeman-lab' has created a pull request for this issue:
https://github.com/apache/spark/pull/3902

> Vector conversion broken for non-float64 arrays
> -----------------------------------------------
>
>                 Key: SPARK-5089
>                 URL: https://issues.apache.org/jira/browse/SPARK-5089
>             Project: Spark
>          Issue Type: Bug
>          Components: MLlib, PySpark
>    Affects Versions: 1.2.0
>            Reporter: Jeremy Freeman
>
> Prior to performing many MLlib operations in PySpark (e.g. KMeans), data are automatically converted to {{DenseVectors}}. If the data are numpy arrays with dtype {{float64}} this works. If data are numpy arrays with lower precision (e.g. {{float16}} or {{float32}}), they should be upcast to {{float64}}, but due to a small bug in this line this currently doesn't happen (casting is not inplace). 
> {code:none}
> if ar.dtype != np.float64:
>     ar.astype(np.float64)
> {code}
>  
> Non-float64 values are in turn mangled during SerDe. This can have significant consequences. For example, the following yields confusing and erroneous results:
> {code:none}
> from numpy import random
> from pyspark.mllib.clustering import KMeans
> data = sc.parallelize(random.randn(100,10).astype('float32'))
> model = KMeans.train(data, k=3)
> len(model.centers[0])
> >> 5 # should be 10!
> {code}
> But this works fine:
> {code:none}
> data = sc.parallelize(random.randn(100,10).astype('float64'))
> model = KMeans.train(data, k=3)
> len(model.centers[0])
> >> 10 # this is correct
> {code}
> The fix is trivial, I'll submit a PR shortly.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org