June 2017: Postdoctoral researcher posts are available at all four institutions.
Links to adverts:
- Strathclyde PDRA (closing date August 31, 2017).
- Oxford PDRA (closing date August 7, 2017).
Closed: Imperial PDRAs (closing date July 20, 2017). Manchester PDRA (closing date July 20, 2017).
For all posts, candidates should be able to work independently but also be willing to collaborate across the project team, and this will involve travel between the four institutions. Candidates will also be involved in direct interaction with external, non-academic, partners, and with public engagement activities. Hence, excellent communication skills and an enthusiasm for outreach are required.
The research at Imperial College will be in the area of mathematical, probabilistic and statistical modelling of human behaviour in urban environments. Predictive policing is one of the major application areas the research associates will investigate. Two PDRAs will work directly with Professor Mark Girolami and will join a vibrant research group and Statistics Section within the Department of Mathematics at Imperial.
The research in Oxford will involve the development and analysis of multilevel Monte Carlo (MLMC) methods applied to the numerical approximation of stochastic PDEs and coupled systems of SDEs. This will include the development of open-source parallel HPC codes on the latest hardware such as GPUs and the Intel Xeon Phi. There will also be some supervision of a PhD student who will be researching the use of MLMC with mixed precision arithmetic.
An ideal candidate would be knowledgeable in MLMC, numerical methods for PDEs, and high performance computing. This specific combination of skills is rare, so a good candidate should have some knowledge in some of these areas, and a clear willingness and ability to learn about the other areas.
The research at Manchester will focus on the numerical linear algebra challenges in the project and on understanding the behaviour in finite precision arithmetic of the algorithms developed. Problems to be considered include the computation of fractional matrix powers and approximation of correlation matrices. Via rounding error analysis, statistical analysis, and numerical experiments, new understanding will be developed of algorithm accuracy, particularly with respect to the use of different precisions of computer arithmetic.
A researcher is sought who has experience in numerical linear algebra and an interest in relevant topics from data science, machine learning, and applied and computational statistics.
At Strathclyde, the research will focus on the development and implementatation of stochastic models for human bevahiour in an urban setting, focusing initially on crime. A suitable candidate will have experience in scientific computation, with skills in one or more of: stochastic modelling (including stochastic differential equations), statistical inference, matrix computations and data science.