University of Strathclyde
Many modern applications produce patterns that change over time.
Important examples arise in mobile telecommunications, on-line trading, smart-metering and on-line social networking. Information such as ‘who called who’, ‘who Tweeted who’,
‘who facebooked who’, and ‘people who bought his book also bought…’ These emerging, data-rich disciplines generate large, high-frequency interaction sequences that demand new computational tools. It is possible to extend standard network concepts, such as paths, walks and geodesics, but any new ideas must respect the arrow of time. Two key observations are that (a) there is an asymmetry in the spread of information around the network (if A meets B and then B meets C, a message may pass A to C, but not vice versa), and (b) using simple aggregate or snapshots to summarize the overall connectivity fails to respect the asymmetry
and systematically overestimates the spread of information. This calls for new ideas in modelling and analysing real-time interactions.
My work has involved designing new computer algorithms that extract key information this type of evolving interaction setting. I have collaborated with colleagues in social media/digital advertising in order to study Twitter networks, and also with colleagues in experimental neuroscience in order to study interaction networks involving brain regions.
The new algorithms allows us address questions such as (a) is the network operating normally?
(b) which parts are most central/vulnerable/peripheral? (c) how will the interactions evolve?
These questions are of great interest in a range of application areas, including smart energy metering, traffic flow, advertising and computer security.
The research has led to new mathematical ideas, computer algorithms, and results where new insights have been extracted from Twitter data and from brain imaging studies.