[
https://issues.apache.org/jira/browse/SPARK-4894?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14277933#comment-14277933
]
Joseph K. Bradley commented on SPARK-4894:
------------------------------------------
+1 for small changes, but occasionally larger ones are needed. The jump to the
spark.ml package might be a good time to make such a switch.
We don't want to try to approach the complexity of Factorie, but it should not
be too hard to support continuous variables. Gaussian NB would surely be the
first generalization to add.
I agree we should think about the API carefully to encourage people to use NB
correctly. The items to decide are:
* Should we only support discrete variables? If so, then there is nothing to
discuss.
* If we support continuous variables (and I would argue we should), then we
must design a good API. No matter what API we design, people may misuse it by
treating a categorical feature as continuous. Documentation is necessary and
helpful, but a good API is important too. That API could take several forms:
** 1 NaiveBayes class, with a Factor type parameter
** multiple NaiveBayes classes for different variable types and/or distributions
** a better way of specifying types, such as enumerations for discrete variables
Basically, to support continuous features, we will need to think about types,
and I agree it should be thought out so it does not become too complex.
> Add Bernoulli-variant of Naive Bayes
> ------------------------------------
>
> Key: SPARK-4894
> URL: https://issues.apache.org/jira/browse/SPARK-4894
> Project: Spark
> Issue Type: New Feature
> Components: MLlib
> Affects Versions: 1.2.0
> Reporter: RJ Nowling
> Assignee: RJ Nowling
>
> MLlib only supports the multinomial-variant of Naive Bayes. The Bernoulli
> version of Naive Bayes is more useful for situations where the features are
> binary values.
--
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
