USI - Email
Sparse spatial random graphs
Host: Prof. Ernst-Jan Camiel Wit
USI Lugano East Campus, room D5.01 // online event
University of Oxford, UK
We present a statistical model to describe spatial random graphs through the graphex process, a construction which exploits the Bayesian nonparametric framework in order to achieve flexibility and interpretability when conducting inference. We provide a number of asymptotic results, namely that the model is able to describe both sparse and dense networks, is equipped with positive global and local clustering coefficients and can achieve both single or double power law degree distributions whose exponents are easily tuned. We present a way to perform posterior inference through MCMC algorithms. Finally, we place our proposal into the more general literature of spatial random graphs, discussing the relations with hyperbolic random graphs, scale-free percolation and sparse latent space models.
Francesca Panero is a visiting researcher at IDSIA and a DPhil candidate in Statistics at the University of Oxford. She is working with Prof. François Caron and Prof. Judith Rousseau on random graphs, exploring both asymptotical properties of models based on graphex processes and new modelling approaches to describe spatial networks. Prior to this, she obtained a BSc and a MSc in Mathematics at the University of Turin and Collegio Carlo Alberto.
INF Newsletter Archive