Universal Approximation Theorem

A foundational theorem in neural network theory stating that a sufficiently large neural network can approximate any continuous function to arbitrary precision. The theorem does not provide quantitative bounds on the required network size.