[
https://issues.apache.org/jira/browse/SPARK-1229?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13975315#comment-13975315
]
Aliaksei Litouka commented on SPARK-1229:
-----------------------------------------
Well... I'm new to Spark so please correct me if I'm wrong. _predict_ methods
declared in ClassificationModel and in KMeansModel accept an RDD\[Vector\] or a
single Vector (not an array!). Each item of the RDD\[Vector\] contains a single
observation for which we want to calculate a prediction, and each item of the
Vector contains a single feature of an observation.
If we need to take a sample of an RDD and feed this sample to any of the train
methods, we can use the _sample_ method instead of _takeSample_ . Actually,
_takeSample_ calls _sample_ internally. However, these methods accept different
parameters. If this is a problem, then I think it's possible to implement
another version of the _sample_ method:
{code}
def sample(withReplacement: Boolean, num: Int, seed: Int): RDD[T]
{code}
Let me know if it makes sense.
> train on array (in addition to RDD)
> -----------------------------------
>
> Key: SPARK-1229
> URL: https://issues.apache.org/jira/browse/SPARK-1229
> Project: Spark
> Issue Type: Story
> Components: MLlib
> Reporter: Arshak Navruzyan
>
> since predict method accepts either RDD or Array for consistency so should
> train. (particularly since RDD.takeSample() returns Array)
--
This message was sent by Atlassian JIRA
(v6.2#6252)
