Temporal Networks

Lead Research Organisation: Imperial College London
Department Name: Dept of Physics

Abstract

The hypothesis for this project is that existing network measures are inappropriate for networks constrained by time or another vertex ordering. The most appropriate measures should reflect this causal structure. Such networks are Directed Acyclic Graphs and are found in many in real data sets. In some contexts (computer science) DAGs play a central role. However in other areas where networks are constrained by time, particularly social contexts such as citation networks, the questions asked are different from those of computer science and the existing DAG tools have not been exploited. This project will adapt existing network tools from different fields and develop new methods including some based on embedding in non-Riemannian (Lorentzian) geometries. The project will investigate these numerically and analytically, on theoretical models of networks and on real large scale data sets with DAG structure.

Publications

10 25 50
publication icon
Vasiliauskaite V (2020) Making communities show respect for order in Applied Network Science

publication icon
Vasiliauskaite V (2019) Social success of perfumes. in PloS one

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/N509486/1 01/10/2016 31/03/2022
1858732 Studentship EP/N509486/1 01/10/2016 31/03/2020 Vaiva Vasiliauskaite
 
Description Network theory is a useful tool to study various interacting entities. For example a friendship network is composed of people as nodes and an edge indicates the two people are friends. In my research, I study a particular type of networks, in which an inherent arrow of time is present---directed acyclic graphs. An example of such network is a citation graph, in which nodes are publications and an edge indicates a citation. Centrality is a network science tool to measure the importance of entities---nodes---with respect to the network. In this research, we showed that some centrality measures assign values of importance to nodes, irrespective of the structure of a network. So, in the context of a citation network, one would find old publications to be important because they are old, not because they are in a structurally important part of the network.
Exploitation Route A natural future extension of this theoretical work is to put it to practice and develop centrality measures which adhere to requirements, put forward in this work.
Sectors Other

 
Description A talk in NetSciX, Tokyo, Japan, 2020 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact A talk, entitled "Making communities show respect for order" was presented in the parallel session talk in NetSciX, Tokyo, Japan in 2020. The talk was received positively, follow-up questions were raised by the audience.
Year(s) Of Engagement Activity 2020
 
Description Lightning talk at NetSci 2018 in Paris, France 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact I delivered a lightning talk on complex networks of scent ingredients and how they influence the success of a perfume based on an on-line database. The lightning talk was presented in the largest conference of the field of complex networks, NetSci, that took place in Paris, France. The lightning talk was delivered in front of all conference participants, more than 500 people. The aim was to introduce a novel tool of studying a complex structure that represents the cultural phenomenon of fragrance and tools to study it using big data. This presentation raised the awareness of new tools and interest into a yet unstudied complex structure.
Year(s) Of Engagement Activity 2018
 
Description Organisation of the Workshop on Advanced Topics in Complex Networks 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach Local
Primary Audience Postgraduate students
Results and Impact On Wednesday the 23rd of May Imperial College's Networks Networking group held a workshop entitled Advanced Topics in Complex Networks which was funded and supported by Imperial College Graduate School's research community fund . The aim of the event was twofold. Firstly, it was an opportunity for PhD students and early career researchers to gain some insight into the "state of the art" mathematical tools being employed by researchers in the area. Secondly, it was to provide a networking opportunity for those in the field. The event was attended by roughly 30-40 PhD students and postdoctoral researchers from Imperial College working on a variety of topics. Attendees came from a range of departments including Ecology (Silwood park campus), Physics, Mathematics and even Economics meaning that the event was truly interdisciplinary. During the morning session, Professor Barahona spoke about "highdimensional data, graph embeddings and multiscale community detection" while Dr Colijn gave a talk entitled "Comparing tree with new metrics and features: applications in evolution and infectious disease." Both talks explored ideas that related to the analysis of large graph data sets and also gave insight into how techniques from machine learning are beginning to creep into network science as well. The talks were followed by a break for lunch which provided an opportunity for the attendees to network with researchers and discuss any questions raised in more detail. The first talk of the afternoon was given by Professor Reimer Kühn, a Professor in mathematics at King's College London. His talk entitled "Spectra of Random Stochastic Matrices and Relaxation in Complex Networks" gave details of the mathematical machinery which can employed of study graph spectra. This is an important topic in applied network science where researchers are often interested in dynamical processes unfolding on graphs. These may include: the spread of disease, models of opinion formation or perhaps the synchronization of stations in power grids. The final talk of the day was from Dr Cornelia Metzig, a postdoctoral researcher at Imperial College who gave a talk concerning her recent research on dynamic network models. This provided an opportunity to learn about the type of questions that practicing researchers in network science are trying to answer. The event raised the importance of networks in a variety of research fields as well as provided an event for a variety of network scientists at Imperial College to gather together and discuss common interests.
Year(s) Of Engagement Activity 2018
URL https://wwwf.imperial.ac.uk/blog/icpostgrad/2018/08/29/workshop-on-advanced-topics-in-complex-networ...
 
Description Presentations in NetSci 2019, Burlington Vermont, USA 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact Parallel session talks on the following projects:
"The longest path in the price model",
"Making communities show respect for order"
were delivered in one of the largest international conferences in the research field. The talks were positively delivered by the field experts, discussions were followed afterwards.
Year(s) Of Engagement Activity 2019