@hvwaldow @tiago

I would say that this statement is reasonable, and I like it!

But from a multi-objective optimization perspective, you would need to find a way to pick "the best" model in terms of accuracy which would lead you to constructing a Pareto set of model's performance in your K-dimensional decision space.

In that case, I assume Tiago would want to either find/construct a model that dominates the Pareto set (based on his prior comments) or would prefer to know with confidence the entire Pareto set, the corresponding Efficient set of models and their relative performance on all the K metrics so he could select one that suits his needs from the set.

Good comment, IMO! 😁