Modeling Temporal Networks

Introduction

Modelling time-evolving networks (such as social, biological and infrastructural) is an open problem that has attracted researchers from a diverse range of fields [1]. The ability to model real networks facilitates our understanding of the nature and the timing of observed evolution, but it may also provide some useful intuition about the future behavior of the network, thereby making valuable predictions. Currently, most existing modelling solutions of temporal networks mimic the real-world networks in terms of certain topological features (number of links, clustering coefficient, degree distribution, connected components, motifs, etc.) but they do not allow us to produce an accurate graph topology [2]. Other approaches focus on temporal link prediction using graph statistics or deep learning methods that fail to provide knowledge about the dynamical process itself [3]. Recently, we have proposed the LG-gen model for modelling temporal networks from a system identification perspective [4]. Our experiments show that many real systems can be well approximated by the LG-gen model.

Project

This project is aimed as one step forward towards a deeper understanding of network evolution processes. Typical questions to be answered include:

  • Modelling: how well the LG-gen model can approximate real unweighted and weighted networks?
  • Link Prediction: what is the performance of LG-gen in comparison to the state-of-art methods such as graph neural networks (GNN)?

In this project, you will study existing graph modelling and link prediction techniques for complex networks and conduct a systematic comparative analysis of their performance on real data.

Requirements

You are in the final stages of your degree in artificial intelligence, computer science, mathematics, electrical engineering or a similar degree. The successful student is expected to have strong programming skills, and have some background in Network Science  You are pragmatic and focused on making things work. Next to technical expertise we value high level of independence and ability to work remotely, communication skills and a results-driven attitude.

To apply

Contact Sergey ShvydunThis email address is being protected from spambots. You need JavaScript enabled to view it.