Model selection is an interesting process, because of overfitting: when you add parameters, are you better modelling the intrinsic noise or a fundamental information that is interesting to you? This is the overarching question, and I find interesting that it depends on the existence of a meta-model, which is usually very implicit.
Model selection is an interesting process, because of overfitting: when you add parameters, are you better modelling the intrinsic noise or a fundamental information that is interesting to you?
This is the overarching question, and I find interesting that it depends on the existence of a meta-model, which is usually very implicit.