Abstract
Information on the organization of a network is usually obtained in terms of the parameter values of a model reproducing the observed topology. However, the parameter choice is often subjective. Here we propose a novel method, based on the Maximum Likelihood principle, to extract a unique, statistically rigorous parameter value from topological data. In this framework, network models turn out to be in general ill-defined or biased; therefore we show a way to define a class of unbiased models. Remarkably, our approach can also be extended in order to extract, only from topological data, the `hidden variables' underlying network organization, making them `no more hidden'. It also solves the problem to correctly randomize a real-world network keeping some of its properties fixed, and allows one to compute averages over the randomized ensemble analytically.
Users
Please
log in to take part in the discussion (add own reviews or comments).