Hyperparameter Tuner

You will work towards implementation of simple Hyperparameter Tuning Framework and this is your high level goal. Your need to produce optimal common.BaseConfig object for each model in models package. To decide if the combination of parameters is optimal please use min and max as fitness function

Framework needs to:

• identify all available models
• determine optimal parameters using each fit function ( min, max )
• return tuning results ( example below )

context:

• It is expected to preserve API of models which can be run using runner.Runner
• Config` object and model functions are externally provided
• Number of models to tune can vary in time
• for purpose of this exercise Model argument data should be considered as Iterable provided to runner.Runner.
• Framework is expected to produce list of optimal sets of parameters for each registered model according to its configuration for each of the fit function ( min and max in this case ). If more than one set of hyperparameters have optimal result of fit function, all of them should be returned. Result can look like:

    - model: x
      results:
      - fit: min
        hyperparameters:
        - a: XXX
          b: YYY
        - a: XXX'
          b: YYY'
      - fit: max
        hyperparameters:
          a: WWW
          b: ZZZ
    - model: y
        results:
      - fit: min
        hyperparameters:
          a: CCC
          b: DDD
      - fit: max
        hyperparameters:
        - a: AAA
          b: BBB
        - a: AAA'
          b: BBB'

constrain:

No modification in common, models and runner should be done in final implementation