Move ML lib data generator files to util/

[shivaram]

The generator classes are out of place in the regression directory --- I've moved them to util to avoid creating another directory (say data or something like that ?)

@mateiz, @etrain, @atalwalkar -- Any other ideas ?

[AmplabJenkins]

Thank you for your pull request. All automated tests for this request have passed.
Refer to this link for build results: http://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/222/

[etrain]

I think this is a fine place to put this - but can we call it "LogisticRegressionDataGenerator"? or something like that?

Also, can we factor the part that produces the RDD into it's own function within the LogisticRegression[Data]Generator object? I think having Data Generation functions available externally that don't rely on writing things to disk will be extremely useful for e.g. unit tests.

Separately, I think it should be pretty easy to set up such a function with an arbitrary input distribution: maybe it could take something like this:
def generateSamples(nsamples: Long, nfeatures: Int, probOne: Double = 0.5, distFun: Random => Double = _.nextGaussian): RDD[Array[Double]]

The last part is probably overkill, though.

[shivaram]

Thanks for the comments. Renamed files to DataGenerator and I also refactored the classes to have methods which can be reused.

[AmplabJenkins]

Thank you for submitting this pull request. Unfortunately, the automated tests for this request have failed.
Refer to this link for build results: http://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/228/

[shivaram]

FYI - the Jenkins build failed because it couldn't reach github.com.

[etrain]

The refactored code looks good to me, thanks Shivavaram!

[mateiz]

Thanks guys, merged this in.