The artificial intelligence of urban dynamics: Neural network modeling of urban structure

Abstract

A neural network, parallel distributed processing model of learning is adapted to represent the self-organizing urban system. The model is trained on a number of cases representing specific functional states of the system, and as a result “learns,” by a process of structural evolution, to recognize the general problem defined implicitly by the set of cases, and to solve it. The learning algorithm approach is based on an explicit distinction between the functional and structural organization of the system; questions such as the structural effects of a functional change are thus addressed directly. Specific results show that very simple models can “learn” to create transportation infrastructure appropriate for a variety of flow requirements, and then distribute flows in a reasonable manner over the network.