As a statistic person, the logic of Param and Params class in Spark ML library confuses me a lot. Since the foundation class PipelineStage extends Params， and transformer/estimator( which means machine learning algorithms to me) are child class of PipelineStage. So Params is not just a simple class represented some machine learning parameters? And the access to some params in pipeline require certain parent requirement. I'm just so lost about how this come out...
Any logical explanation for Param? Thanks a lot! This really confuses me.