Figure 4:

The property of universality means that neural networks can represent any continuous function. The neural network shown here represents a hypothetical system to take a photographic image of a patient and render a prediction of their Cormack-Lehane view at intubation. (Not all nodes and connections are illustrated, as the input and hidden layers would each contain several thousand nodes. More pragmatic network topologies can be applied to visual recognition problems than the general case shown here.)