Graph Signal Processing on Dynamic Graphs

This PhD project is inline with EPSRC priority on artificial intelligence and robotics, aimed at developing novel learning methodologies for large scale networks. Networks occur frequently in real datasets such as social networks, e-commerce and financial transactions, in addition to various scientific and technological areas. With the rapidly increasing demand for analysing complex network-structured data, network science and network analysis has become one of the hottest research topics in the field of statistical data science and big data analytics. However, most statistical network analysis methods have been targeted at static networks, and the understanding of dynamic networks, a type of network constantly changes over time, is relatively poor. In view of the clear importance of dynamic networks, this project will build a theoretical framework for dynamic network analysis. It will involve metrics design, mathematical modeling and statistical analysis, as well as experimental validation. The research and validation will be carried out on publicly available datasets of human mobility and social networks. The ultimate goal of this work is to benefit the greatest number of people, so it will be utilized for several real-life problems, such as prediction of the spread of infectious diseases, analysis of social networks for recommendations (e.g. product or music recommendations) and prediction of demands in financial transactions.