[
https://issues.apache.org/jira/browse/SPARK-9075?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14628671#comment-14628671
]
Joseph K. Bradley commented on SPARK-9075:
------------------------------------------
I agree it's not a bug, but it is a requirement which we could relax. The
requirement is pretty reasonable IMO. If you have numExamples <
maxCategoriesPerFeature, then you will know nothing about some of your feature
values and will have to make arbitrary decisions about how to split that
feature.
Possible solutions include:
* Ignoring categorical features if we cannot collect enough info about them
(This is not a usual practice in popular ML libraries, so I would not recommend
it.)
* For high-arity categorical features, we could create a small number of bins
which randomly group categories. This would be a more reasonable solution, but
I think it's low-priority since random groupings of many categories would
provide very little information.
In the meantime, I'd recommend we improve the error message to tell the user to
remove a feature X when there is too little data to make use of X. I'll close
this, but will create & link a JIRA for improving the error message.
> DecisionTreeMetadata - setting maxPossibleBins to numExamples is incorrect.
> ----------------------------------------------------------------------------
>
> Key: SPARK-9075
> URL: https://issues.apache.org/jira/browse/SPARK-9075
> Project: Spark
> Issue Type: Bug
> Components: MLlib
> Affects Versions: 1.4.0
> Reporter: Les Selecky
> Priority: Minor
>
> In
> https://github.com/apache/spark/blob/master/mllib/src/main/scala/org/apache/spark/mllib/tree/impl/DecisionTreeMetadata.scala
> there's a statement that sets maxPossibileBins to numExamples when
> numExamples is less than strategy.maxBins.
> This can cause an error when training small partitions; the error is
> triggered further down in the logic where it's required that
> maxCategoriesPerFeature be less than or equal to maxPossibleBins.
> Here's the an example of how it was manifested: the partition contained 49
> rows (i.e., numExamples=49 but strategy.maxBins was 57.
> The maxPossibleBins = math.min(strategy.maxBins, numExamples) logic therefore
> reduced maxPossibleBins to 49 causing the "require(maxCategoriesPerFeature <=
> maxPossibleBins" to throw an error.
> In short, this will be a problem when training small datasets with a feature
> that contains more categories than numExamples.
> In our local testing we commented out the "math.min(strategy.maxBins,
> numExamples)" line and the decision tree succeeded where it had failed
> previously.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
