School of Electronics, Electrical Engineering and Computer Science
& ECIT Global Research Institute
Proposed Project Title: Future Directions in Graph Analytics
Principal Supervisor: Hans Vandierendonck
Second Supervisor: TBC (pending project details)
In the era of big data, numerous data analytics problems require the processing of large graphs. These problems, often referred to as “graph analytics” are challenging to process with high computational efficiency. Their scaling in distributed and disk-based processing systems is significantly hindered by frequent communication and by the irregular structure of scale-free graphs. During the past three years, dr Hans Vandierendonck and colleagues at Queen’s have made significant findings in relation to load balancing parallel graph analytics and improving memory locality of graph analytics.
We are seeking excellent PhD candidates to pursue this research along the following strands:
- Data structures for storing graphs and supporting efficient graph analytics for static and dynamic graphs; and understanding at a fundamental level the cause of the performance differences between “top-down” or “forward” graph traversal versus “bottom-up” or “backward” traversal.
- Vectorisation of graph analytics, including auto-vectorisation in a domain-specific language for graph analytics
- Addressing limitations of the Pregel programming model: extending the model to support algorithms such as union-find that do not fit the model; supporting asynchronous computation in the programming model
- Developing graph algorithms using transprecision computing (the act of adapting the precision of floating-point calculations at runtime) and the support for transprecision in the programming model and runtime system.
Prospective PhD students with an interest in these and other problems in graph analytics are welcome to contact dr Hans Vandierendonck at email@example.com, home page: http://www.eeecs.qub.ac.uk/~H.Vandierendonck/
Supervisor Name: Hans Vandierendonck Tel: +44 (0)28 9097
QUB Address: Computer Science Building, 18 Malone Road, Belfast BT9 5BN Email: firstname.lastname@example.org