How to Create Admissible Set of Functions?

Creating an admissible set of functions involves managing complexity while ensuring a small VC dimension. Classical learning theory typically provides a predefined set of functions, which can be inadequate as the challenge lies in deriving these functions from a continuous set. The goal is to construct a finite-dimensional admissible set that encapsulates good functions useful for learning. The process requires understanding properties of training data and their average behavior across functions, representing a kind of variance. It is crucial to ask the right questions related to the problem domain to distinguish between trivial and meaningful functions. Intelligence in this context involves identifying specific predicates that relate to the problem at hand, thereby enhancing the modeling process.