This studentship has been developed by the University of Liverpool in partnership with GCHQ.
Over recent years, Deep Learning has been demonstrated to offer impressive performance across a range of application domains. In part, this is because of its ability to exploit modern many-cored processors (e.g., GPUs).
At its heart, Deep Learning makes use of stochastic gradient ascent, an approach that is known to struggle in the context of both local maxima and (the more numerous) saddle-points, degrading performance possible using a fixed dataset. Techniques have been developed that replace the stochastic gradient ascent with numerical Bayesian methods typified by Markov chain Monte Carlo (MCMC). Unfortunately, while such use of MCMC makes it theoretically possible to always find the global maximum, MCMC is so slow that this is hard to achieve in practice. This has restricted the extent to which MCMC has been used in the context of Deep Learning.
Recent work by the University of Liverpool has developed a variant of the Sequential Monte Carlo (SMC) sampler, as a generic alternative to MCMC. SMC samplers appear to offer the same global optimisation capability as MCMC but with substantially reduced computational cost. This comes about, in part, because SMC samplers are inherently parallel, making it possible to exploit, for example, GPUs.
Deep Learning achieves parallelism by distributing the processing across the data. SMC samplers achieve parallelism by distributing the consideration of the many hypotheses that the data could imply to be true. This studentship will begin by investigating the extent to which a combination of Deep Learning and SMC samplers can be parallelised by simultaneously considering parallelism across the data and the hypotheses. The studentship will then go on to undertake research that is only possible as a result of the existence of fast Bayesian Deep Learning (e.g., related to robust outlier detection).
This project is part of the EPSRC Funded CDT in Distributed Algorithms: The What, How and where of Next-Generation Data Science. https://www.liverpool.ac.uk/research/research-at-liverpool/research-themes/digital/cdt-distributed-algorithms/
The University of Liverpool is working in partnership with the STFC Hartree Centre and other industrial partners from the manufacturing, defence and security sectors to provide a 4 year innovative PhD training course that will equip over 60 students with the essential skills needed to become future leaders in data science, be it in academia or industry.
Every project within the centre is offered in collaboration with an Industrial partner who as well as providing co-supervision will also offer the unique opportunity for students to access state of the art computing platforms, work on real world problems, benchmarking and data. Our graduates will gain unparalleled experiences working across academic disciplines in highly sought-after topic areas, answering industry need.
As well as learning from academic and industrial world leaders, the centre has a dedicated programme of interdisciplinary research training including the opportunity to undertake modules at the global pinnacle of Data science teaching. A large number of events and training sessions are undertaken as a cohort of PhD students, allowing you to build personal and professional relationships that we hope will lead to research collaboration either now or in your future.
The learning nurtured at this centre will be based upon anticipation of the hardware recourses arriving on desks of students after they graduate, rather than the hardware available today.
To apply for this Studentship please submit an application for an Electrical Engineering PhD via our online platform (https://www.liverpool.ac.uk/study/postgraduate-research/how-to-apply/
) and provide the studentship title and supervisor details when prompted. Should you wish to apply for more than one project, please provide a ranked list of those you are interested in.
For a full list of the entry criteria and a recruitment timeline (including interview dates etc), Please see our website https://www.liverpool.ac.uk/research/research-at-liverpool/research-themes/digital/cdt-distributed-algorithms/
For informal enquires please contact Prof Simon Maskell [email protected]
or [email protected]
Professor Vassil Alexandrov - STFC Hartree Centre
I’ve worked in high performance computing (HPC), data and computational science for a long time, with a fulfilling career spanning 18 years and 5 countries! I’ve also published over 130 papers in journals and at international conferences and workshops. I’m excited to be a supervisor so that I can pass on my knowledge and experience to the next generation of young people who will develop research projects in exciting areas of HPC and data science.
During my career, I have supervised 31 PhD students to successful completion of their PhD studies across a variety of computational themes and areas, and been a Programme Director of 3 MSc programmes. I am a member of the Editorial Board of the Journal of Computational Science (JOCS) and Editor of Mathematics and Computers in Simulation journal.
Beginning in Russia, I achieved an MSc degree in Applied Mathematics from Moscow State University, followed by a PhD degree in Parallel Computing from Bulgarian Academy of Sciences. I have also previously held positions at the University of Liverpool, UK, the University of Reading, UK, and Monterrey Institute of Technology and Higher Education (ITESM), Mexico.
In 2019 I was appointed as Chief Science Officer at the Science and Technology Facilities Council (STFC) Hartree Centre in the UK. Previously I was an ICREA Research Professor in Computational Science at Barcelona Supercomputing Centre, Spain.
I have a lot of experience in stochastic modelling, Monte Carlo methods and algorithms, parallel algorithms and scalable algorithms for extreme scale computing, e.g. for large-scale systems and applications. My long-term expertise in Monte Carlo means I am particularly interested in seeing how we can further speed up these simulations.
Currently, mathematics-led innovation is clearly indispensable in advancing key scientific areas, as well as powering methods and algorithms enabling to discover global properties of data.