...

An important part of the Flink ML library infrastructure is the APIs to create, set and get parameters of machine learning stages (e.g. Transformer, Estimator). Currently Flink ML library provides this functionality through the public methods of WithParams (an interface), ParamInfo, ParamInfoFactory and Params (three classes). This doc proposes to simplify the Flink ML library infrastructure by reducing the number of classes as well as the number of public methods on those classes, and still delivering the same functionality.

The goal is to make of this FLIP is to increase developer velocity by making Flink ML library easier to maintain and understood by developersuse.

Analysis of the existing parameter-related interface and classes

...