In machine learning, an admissible set of functions is a set with small capacity or diversity, and a good function can be found within this set. VC theory refers to a way of measuring the complexity of an admissible set, which is important when selecting a function for a machine to use in making predictions.