Font Size:
A latent space model for multidimensinoal networks
Last modified: 2017-05-08
Abstract
A multidimensional network is a collection of networks (views). The views are dened on a
constant node set but, potentially, on dierent edge sets. This kind of structure well describes either multiple characteristics of a group of units or a phenomenon changing over time. We present a latent space approach to model binary multiplex data, where the probability of having a linked dyad in a view is modelled as a function of the network's connectivity and of the distance between its nodes in a latent space. A constant node set allows to represent the nodes in a single latent space, common for the whole multidimensional networks. We adopt a hierarchical Bayesian approach and use MCMC inference to estimate the parameters. The special case with varying node sets will be discussed and an application on real data will be presented.
constant node set but, potentially, on dierent edge sets. This kind of structure well describes either multiple characteristics of a group of units or a phenomenon changing over time. We present a latent space approach to model binary multiplex data, where the probability of having a linked dyad in a view is modelled as a function of the network's connectivity and of the distance between its nodes in a latent space. A constant node set allows to represent the nodes in a single latent space, common for the whole multidimensional networks. We adopt a hierarchical Bayesian approach and use MCMC inference to estimate the parameters. The special case with varying node sets will be discussed and an application on real data will be presented.