Understanding and modelling the robustness of complex networks for varying degrees of structural information

Abstract

Complex networks provide a framework for modelling and understanding the behaviour of a wide variety of real world systems, and one network property of interest is robustness. Here we explore several different understandings of network robustness, developing new models and methods in order to predict network robustness and identify when existing predictive methods fail. We begin by introducing a new analytic model for predicting the robustness of ecological species interaction networks which also provides novel insights into what structures make ecological networks robust. We go on to relate various information theoretic measures to robustness on more generic networks, determining what we can know about network robustness given certain levels of structural information. Statistical models which incorporate a variety of information about network structure in order to predict the robustness of networks are tested and compared. Finding limitations in the capability of all existing methods to accurately predict robustness, we develop a new method which outperforms existing methods in either accuracy or computational efficiency. Finally, we identify the network structures which lead to poor predictions of robustness, informing us of the type of networks which require more sophisticated methods in order to be well predicted in the future.

Date of Award	3 Oct 2023
Original language	English
Awarding Institution	University of Bristol
Supervisor	Stephen R Wiggins (Supervisor) & Karoline Wiesner (Supervisor)