Universal Approximation Theorem

The theorem stating neural networks with one hidden layer can approximate any continuous function.

This concept is essential for understanding machine learning fundamentals and forms a key part of modern AI systems.