A comparative analysis of specific spatial network topological models

Creating ensembles of random but "realistic" topologies for complex systems is crucial for many tasks such as benchmark generation and algorithm analysis. In general, explanatory models are preferred to capture topologies of technological and biological complex systems, and some researchers claimed that it is largely impossible to capture any nontrivial network structure while ignoring domain-specific constraints. We study topology models of specific spatial networks, and show that a simple descriptive model, the generalized random graph model (GRG) which only reproduces the degree sequence of complex networks, can closely match the topologies of a variety of real-world spatial networks including electronic circuits, brain and neural networks and transportation networks, and outperform some plausible and explanatory models which consider spatial constraints.