This PR provides a draft for the restructured DoEStrategy.
The following things changed:
The objective is now criterion to prevent confusion with the objectives of the output features.
The former objectives like DOptimality are not anymore enums, but classes, for example DOptimalityCriterion. This allows for the clean implementation of criterion specific attributes.
The formula is now anymore a direct attribute of the DoEStrategy but instead an attribute of the criterion.
When the functional part of the DoE module is refactored, we should have a mapper which takes the criterion and domain and returns a functional criterion class which provides then the actual objective funtion for the actual ipopt optimizer.
On the long run, we can then also provide a common base class on the data model level for both the DoEStrategy and the SpaceFillingStrategy. In case of the latter one the only allowed criterion will then be the SpaceFillingCriterion which has no formula attribute.
This PR provides a draft for the restructured
DoEStrategy.The following things changed:
objectiveis nowcriterionto prevent confusion with the objectives of the output features.DOptimalityare not anymore enums, but classes, for exampleDOptimalityCriterion. This allows for the clean implementation of criterion specific attributes.DoEStrategybut instead an attribute of thecriterion.When the functional part of the DoE module is refactored, we should have a mapper which takes the criterion and domain and returns a functional criterion class which provides then the actual objective funtion for the actual ipopt optimizer.
On the long run, we can then also provide a common base class on the data model level for both the
DoEStrategyand theSpaceFillingStrategy. In case of the latter one the only allowed criterion will then be theSpaceFillingCriterionwhich has noformulaattribute.@Osburg: What do you think?