You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@spark.apache.org by "Mathieu DESPRIEE (JIRA)" <ji...@apache.org> on 2017/06/29 19:35:00 UTC

[jira] [Created] (SPARK-21257) LDA : create an Evaluator to enable cross validation

Mathieu DESPRIEE created SPARK-21257:
----------------------------------------

             Summary: LDA : create an Evaluator to enable cross validation
                 Key: SPARK-21257
                 URL: https://issues.apache.org/jira/browse/SPARK-21257
             Project: Spark
          Issue Type: New Feature
          Components: ML
    Affects Versions: 2.1.1
            Reporter: Mathieu DESPRIEE


I suggest the creation of an LDAEvaluator to use with CrossValidator, using logPerplexity as a metric for evaluation.

Unfortunately, the computation of perplexity needs to access some internal data of the model, and the current implementation of CrossValidator does not pass the model being evaluated to the Evaluator. 
A way could be to change the Evaluator.evaluate() method to pass the model along with the dataset.




--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

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