The present paper is about viewing a given network as a realization of a network which belongs to a certain set of networks, called an ensemble. It is comparable to viewing a population as being generated from a superpopulation.
The aim of the present paper is to generate a set of networks that resembles a given network in that they have certain aspects in common that are considered important. These networks can be used to compute certain characteristics, and see how they vary. In a sense, the goal of this method is comparable to bootstrapping, although it achieves this in a different way. The idea of an ensemble of networks is inspired by statistical mechanics.