[SPARK-29377][PYTHON][ML] Parity between Scala ML tuning and Python ML tuning

[huaxingao]

What changes were proposed in this pull request?

Follow Scala ml tuning implementation

• put leading underscore before python ValidatorParams to indicate private
• add _CrossValidatorParams and _TrainValidationSplitParams
• separate the getters and setters. Put getters in _XXXParams and setters in the Classes.

Why are the changes needed?

Keep parity between scala and python

Does this PR introduce any user-facing change?

add CrossValidatorModel.getNumFolds and TrainValidationSplitModel.getTrainRatio()

How was this patch tested?

Add doctest

[SparkQA]

Test build #111911 has finished for PR 26057 at commit 8a5615b.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):
• class _ValidatorParams(HasSeed):
• class _CrossValidatorParams(_ValidatorParams):
• class CrossValidator(Estimator, _CrossValidatorParams, HasParallelism, HasCollectSubModels,
• class CrossValidatorModel(Model, _CrossValidatorParams, MLReadable, MLWritable):
• class _TrainValidationSplitParams(_ValidatorParams):
• class TrainValidationSplit(Estimator, _TrainValidationSplitParams, HasParallelism,
• class TrainValidationSplitModel(Model, _TrainValidationSplitParams, MLReadable, MLWritable):

[srowen]

Will these changes affect users who try to extend these classes much or at all?

[huaxingao]

I don't think adding leading underscore will affect users to extend these classes. The single leading underscore before a class name is only a weak indicator for internal usage. It doesn't enforce privacy.

[zhengruifeng]

retest this please

[SparkQA]

Test build #111974 has finished for PR 26057 at commit 8a5615b.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):
• class _ValidatorParams(HasSeed):
• class _CrossValidatorParams(_ValidatorParams):
• class CrossValidator(Estimator, _CrossValidatorParams, HasParallelism, HasCollectSubModels,
• class CrossValidatorModel(Model, _CrossValidatorParams, MLReadable, MLWritable):
• class _TrainValidationSplitParams(_ValidatorParams):
• class TrainValidationSplit(Estimator, _TrainValidationSplitParams, HasParallelism,
• class TrainValidationSplitModel(Model, _TrainValidationSplitParams, MLReadable, MLWritable):

[viirya]

Do we need to add since for setters? Some like setNumFolds have it, but some do not.

[huaxingao]

If the original file has @SInCE, I have it too. Some of the getters/setters don't have @SInCE in the original file, so I don't have i too.
I will check the history to find out which release starts to have these getters/setters, and add @SInCE for each of the getters/setters.

[huaxingao]

I checked the history, the @since was removed in #12020. I guess the reason is that in CrossValidator, the getters/getters have @since 1.4, but in TrainValidationSplit, the getters/getters have @since 2.0. When putting the common getters/setters in ValidatorParams, neither @since 1.4 nor @since 2.0 is quite right, so @since was removed. I am facing the same problem. Shall I have @since 2.0? or just not adding back the @since?

[srowen]

@since 2.0 is at least accurate, as both have existed since that version. One existed before, sure, but this is maybe not a big deal, as it won't matter much at all to know it was in 1.x.

[SparkQA]

Test build #112000 has finished for PR 26057 at commit 2a23e08.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[zhengruifeng]

merged to master, thanks all

[huaxingao]

@zhengruifeng @viirya @srowen Thanks all for your help!