This PhD project is part of the CDT in Distributed Algorithms: The What, How and where of Next-Generation Data Science.
The University of Liverpool’s Centre for Doctoral Training in Distributed Algorithms (CDT) is working in partnership with the STFC Hartree Centre and 20+ external partners from the manufacturing, defence and security sectors to provide a 4-year innovative PhD training programme that will equip up to 60 students with: the essential skills needed to become future leaders in distributed algorithms; the technical and professional networks needed to launch a career in next generation data science and future computing; and the confidence to make a positive difference in society, the economy and beyond.
Working with NSG, one of the worlds largest manufacturers of glass and glazing products, this successful PhD candidate will investigate the use of SMC samplers and High-performance computing with a focus on representing the uncertainty in the context of an efficient and more accurate Multi-Output Gaussian Process as used in Bayesian Optimisation.
Simulation techniques exist to assess how feasible it is for existing production processes to be used to make a new glazing part for a car. While instances of these simulations have been run extensively in the past, it is unlikely that the new glazing part will be identical to one that has been assessed previously. At present, optimising the parameters of the simulation to provide such assessments of new parts is labour-intensive and iteratively re-running the simulation is time-consuming. It would be highly advantageous if it were possible to identify similar historic glazing parts, identify the associated inputs to the simulations, interpolate between the associated pre-existing simulation outputs and so predict the feasibility for a new part. It would also be desirable to recommend the inputs to the simulations that, when run, would be most likely to reduce the uncertainty associated with any such interpolation. In fact, there are a variety of simulation techniques each of which has been historically applied to a different set of glazing parts and real-world data that have been used to validate these simulations. These historically simulated and real-world outputs agree in general, but it is very important to understand the disparities, eg when the simulations predict something is feasible and, in reality, it is not. It is therefore important for any analysis to consider the similarities between these outputs as a function of the inputs.
Multi-Output Gaussian Processes can represent the uncertainty associated with interpolation and extrapolation across the historic simulation and real-world outputs. Given the articulation of uncertainty, Bayesian Optimisation can be used to tackle this problem of recommending which simulation to run next given a generic objective of understanding the historic data or a specific objective of assessing feasibility of a new glazing part. However, Gaussian Processes require a mathematical model that defines similarity, ie a way to look at the shape of two glazing parts and assess whether the parameters pertinent to assessing feasibility are similar. Estimating the parameters of a mathematical model for similarity is an example of hyper-parameter estimation. Hyper-parameter estimation can be solved using existing numerical Bayesian techniques typified by Markov Chain Monte Carlo (MCMC). MCMC is inherently designed for a single processor and is therefore likely to be slow in this context. Thankfully, there is an alternative technique, the Sequential Monte Carlo (SMC) sampler, that makes it possible to exploit parallel processing to make the same inferences as MCMC in a fraction of the time. This ensures that the scheduling of any simulations can exploit the same High-Performance Computing as the simulations themselves.
We welcome UK nationals and international applicants from countries where NSG has a significant manufacturing presence – individuals from the following countries are eligible to apply: UK, EU, USA, Mexico, Brazil, Argentina, Chile, Malaysia, Japan & Vietnam.
Due to the competitive nature of this opportunity, we are receiving a high number of applications and so only applications and enquiries from students from the eligible countries listed above will be considered and should expect to receive a reply.
Students are based at the University of Liverpool and part of the CDT and Signal Processing research community. Every PhD is part of a larger research group which is an incredibly social and creative group working together solving tough research problems. Students have 2 academic supervisors and an industrial partner who provides co-supervision, placements and the opportunity to work on real world challenges. In addition, students attend technical and professional training to gain unparalleled expertise to make a difference now and in the future.
This studentship is due to commence 1 October 2022 (Covid-19 Working Practices available).
To apply please contact the supervisors Dr Pete Green (email@example.com) and Dr Xiaowei Huang (Xiaowei.Huang@liverpool.ac.uk) in the first instance or visit the CDT website for Director, Student Ambassador and Centre Manager details: https://www.liverpool.ac.uk/distributed-algorithms-cdt/
Visit the CDT website for application instructions, FAQs, interview timelines and tips.