# How are two layer feed-forward neural networks universal?

Across my studies I have noticed the following statement in my Subject Guide; namely, that two-layer feed-forward neural networks using the sigmoidal activation function are universal. My question is how are the networks 'universal' and what does 'universal' actually mean in this instance?

I think they are talking about the universal approximation theorem which states that given a continuous function $f$ over an n-dimensional input vector $\vec{x}$, then a neural network with a single hidden layer can approximate $f(\vec{x})$ arbitrarily closely.