[jira] [Created] (SPARK-20673) LDA `optimizer` do not really support case insensitive

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

zhengruifeng created SPARK-20673:
------------------------------------

             Summary: LDA `optimizer` do not really support case insensitive 
                 Key: SPARK-20673
                 URL: https://issues.apache.org/jira/browse/SPARK-20673
             Project: Spark
          Issue Type: Bug
          Components: ML
    Affects Versions: 2.2.0
            Reporter: zhengruifeng
            Priority: Minor


The declaration of {{optimizer}} in {{LDA}} suggest that it should be case insensitive:
{code}  
final val optimizer = new Param[String](this, "optimizer", "Optimizer or inference" +
    " algorithm used to estimate the LDA model. Supported: " + supportedOptimizers.mkString(", "),
    (o: String) =>
      ParamValidators.inArray(supportedOptimizers).apply(o.toLowerCase(Locale.ROOT)))
{code}

However, it do not support uppercase:
{code}
scala> val lda = new LDA().setOptimizer("Em")
lda: org.apache.spark.ml.clustering.LDA = lda_dde23ccc5792

scala> lda.getOptimizer
res18: String = Em

scala> lda.fit(dataset)
scala.MatchError: Em (of class java.lang.String)
  at org.apache.spark.ml.clustering.LDAParams$class.getOldOptimizer(LDA.scala:351)
  at org.apache.spark.ml.clustering.LDA.getOldOptimizer(LDA.scala:809)
  at org.apache.spark.ml.clustering.LDA.fit(LDA.scala:898)
  ... 54 elided
{code}



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

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