SAMBa welcomes you to the second annual SAMBa Summer Conference, which will take place on Monday 25th and Tuesday 26th of June 2018. This page will be updated in the lead up to, and during the event, with plans (as they become concrete), titles, abstracts and slides.
Registration is now closed. If you would still like to attend, please email Cameron Smith or Elizabeth Gray as soon as possible.
The conference is an opportunity for SAMBa students to showcase the vast range of statistical applied mathematics to members of the department, outside the department and at other Universities. SAMBa has students working across the spectrum of statistical applied mathematics, including statistics, probability, numerical analysis, fluid dynamics, mathematical biology, machine learning, high performance computing to name but a few. The conference is organised by the students, and contains talks and presentations by SAMBa students, as well as external speakers and students from other departments and institutions.
Organisers: Elizabeth Gray (firstname.lastname@example.org) and Cameron Smith (email@example.com)
Location: Department of Mathematical Sciences, University of Bath, BA2 7AY
If you have any questions about the conference, please feel free to email either Elizabeth or Cameron.
|Monday 25th June||Tuesday 26th June|
|09:00||Joint Session with CSCT |
Sponsored by Syngenta
|12:00||Photo and Lunch
(Photo: Outside 6W)
(Lunch: 4W Atrium)
|8 Minute Talks
|17:00||Wine and Networking
(The Stable, Bath)
Invited Speaker: Prof. David Silvester (Manchester University)
Title: Robust preconditioning of stochastic Galerkin approximation of nearly incompressible elasticity.
The locking of finite element approximations when solving nearly incompressible elasticity problems is a significant issue in the computational engineering world. In this talk we consider the linear elasticity problem with an uncertain spatially varying Young's modulus. The uncertainty is modelled with a finite set of parameters with prescribed probability distribution. We introduce a novel three-field mixed variational formulation of the PDE model and discuss its approximation by stochastic Galerkin mixed finite element techniques. The S-IFISS software used for computation is available online.
Student Speaker: Matt Griffith (Cohort 3 aligned)
Title: Raising the Roof: Using the Unified Model to simulate the lower thermosphere. Results and validation.
Forecasting weather in the lower thermosphere (85 - 120 km) is of particular interest due to its impact on spacecraft re-entry and radio communications. To this end, we aim to extend the current 85 km upper boundary on the Met Office's Unified Model (UM) to a height of around 120 km. Thus, we shall raise the roof on current numerical weather prediction and pave the way for the development of a whole atmosphere model.
This region, however, has proven to cause particular difficulties for the UM. Thus, extended UM simulations were performed to assess the model anomalies directly. With these model runs, the radiation scheme was discovered to be a significant contributing factor causing the model to crash. In particular the lack of consideration of non-Local Thermodynamic Equilibrium (LTE) effects in the thermosphere leads to anomalous shortwave radiative heating.
The inclusion of non-LTE effects is still a work in progress. Thus, in order to circumvent this problem, we shall replace the radiation scheme by nudging towards a climatological temperature structure above 70 km. With this in place, we look to validate the model's accuracy in the lower thermosphere by comparison to data.
In particular, we shall focus on accurately depicting the reversal of the zonal jets, forced by gravity waves (GWs). In order to do this, tuning of the GW forcing schemes is required. We can then make a comparison with available radar and satellite data for the GW profile. We present and explain this data, and describe how it will be used to validate and enforce accuracy on the extended model.
Student Speaker: Tom Pennington (Cohort 3)
Title: Statistics and Geometry: A Mat(c)h made in Heaven.
A statistical model likelihood can be viewed as a map from parameter space to the set of probability distributions on data space. Studying geometry on either space can provide insight for a range of problems, including speech recognition, computer vision and Bayesian inference. My talk will present a whirlwind tour of geometrically inspired methods for statistics and machine learning.
Invited Speakers: Dr. Kirsty Hassall and Dr. Alice Milne (Rothamsted Research)
Title: Understanding variation in data: a brief tour through applications in agriculture.
Data are being generated in ever increasing amounts in many different applications at all spatial scales. Agriculture is no exception. However, it is often far from clear how data can be related back to scientific questions of interest. In this talk, we will describe a number of examples, where mathematics and statistics have been fundamental in the translation from "data" to "information". Examples will include the design of appropriate sample schemes to enable us to elucidate relationships between biological variables, how we can use sensor data to inform management zones in farmers' fields, using signal processing methods to understand complex variation and using Bayes Net modelling to operationalize the notion of soil health.
Student Speaker: Aoibheann Brady (Cohort 2)
Title: Attribution of large scale drivers for flood risk & the problem of causality
In this talk, we discuss the attribution of trends in peak river flows to large-scale climate drivers such as the North Atlantic Oscillation (NAO) and the Eastern Atlantic (EA) Index. We focus on a set of near-natural "benchmark" catchments in the UK in order to detect trends not caused by anthropogenic changes, and aim to attribute trends in peak river flows to some climate indices. To improve the power of this approach compared with the classical method of at-site testing, we propose modelling all stations together in a Bayesian framework.
This approach leads to the detection of a clear countrywide time trend. Additionally, the EA appears to have a considerable association with peak river flows, particularly in the south east of the UK. However, this is an association only, not necessarily a causal link. We discuss the problem of assessing for causal relationships in environmental studies such as this, and present potential solutions for assessing for causality in these cases.
Student Speaker: Abigail Verschueren (Cohort 4 aligned)
Title: Joint modelling of longitudinal and survival data.
In clinical trials, it is often beneficial to stop a trial early. Group sequential trial designs provide stopping rules that can, on average, result in early stopping or recruitment of less patients. Many clinical trials concern data where the primary endpoint is survival, in which case we expect a long delay in observing the response. However, suppose another measurement can be made that acts as a predictor for survival, called a "biomarker", then this can be used to create an effective stopping rule.
In this talk I will introduce group sequential clinical trials and discuss the joint modelling framework for longitudinal and survival data that describes the relationship between the two processes. I then explain how the joint model is extended to group sequential trials and how this allows for earlier stopping.
Invited Speaker: Dr. Amanda Turner (Lancaster University)
Title: One-dimensional scaling limits in a planar random growth model.
Planar random growth processes occur widely in the physical world. Examples include diffusion-limited aggregation (DLA) for mineral deposition and the Eden model for biological cell growth. One of the curious features of these models is that although the models are constructed in an isotropic way, scaling limits appear to be anisotropic. In this talk, we construct a family of models in which randomly growing clusters can be represented as compositions of conformal mappings. We are able to show rigorously that for certain parameter choices, the scaling limits are anisotropic and we obtain shape theorems in this case. This contrasts with earlier work on related growth models in which the scaling limits are shown to be growing disks.
Student Speaker: Emma Horton (Cohort 3 aligned)
Title: Stochastic Analysis of the Neutron Transport Equation.
The neutron transport equation (NTE) describes the net movement of neutrons through an inhomogeneous fissile medium, such as a nuclear reactor. The nuclear fission processes in such reactors can be realised as branching processes, whose linear semigroups solve the NTE. In this talk I will describe the dynamics of the associated branching process, and give some results regarding the existence of the leading eigenvalue and eigenfunction of the NTE. I will also talk about different simulation techniques of the neutron transport process, which allow us to obtain these quantities numerically.
Student Speaker: Andrea Lelli (Cohort 2)
Title: Mixing Time for Subcritical Random-Cluster Dynamics.
Random-Cluster model can be seen as a generalisation of the independent percolation model, where the introduction of a parameter on the number of the clusters makes the status of each edge depend on the configuration of the entire graph. Because of this global dependence, to the present, still very little is known on the mixing time of the dynamics of the Random-Cluster model, namely the Glauber dynamics. We prove an upper bound for the mixing time of the Glauber Dynamics of the subcritical Random-Cluster on any vertex-transitive lattice.
First year SAMBa students give quick-fire insights into their upcoming research or a previous project.
William Graham: Physical models for designation of nitrate vulnerable zones.
Teo Deveney: Gaussian process latent variable models for images.
Allen Hart: Can neural networks tell you how you feel?
Yyanis Johnson-Llambias: New asymptotic theories for singular 3D and time dependent hydrodynamics.
Thomas Finn: Multi-particle diffusion limited aggregation.
Eleanor Barry: Systematic sclerosis & cancer.
Shaunagh Downing: Marine seismic imaging.
Kevin Olding: Mini flash crashes.
Lizhi Zhang: Statistical inference on network data.
SAMBa students present some maths (research related or otherwise!) in a series of short talks.
Tom Smith: Some challenges with using non-systematic records.
Beth Boulton: Now you've all got worms!
Owen Pembery: Listening is harder when it's noisy.
Matthias Klaar: Estimating π with matches.
Cameron Smith: Reaction-diffusion systems.
Kgomotso Morupisi: A non-smooth model of glacial cycles.
Note that a * denotes students who are away at conferences or on industrial placements. Numbers refer to the poster-board number.
1. Jack Betteridge: Solving PDEs using the discontinuous Galerkin method in DUNE and Firedrake.
2. Aoibheann Brady: Detection & attribution of large-scale drivers for flood risk in the UK.
3. Federico Cornalba: A regularised Dean-Kawasaki model: derivation and analysis.
4. Elizabeth Gray: Dealing with preferentially sampled data.
5. Nadeen Khaleel: Bayesian inference for crime incedence in the USA.
6. Amelie Klein: Sound classification using machine learning.
7. Robbie Peck*: Phase II/III programs: optimal setup and decision making.
8. Lizzi Pitt: Optimising 'First In Human' trials through dynamic programming.
9. Adwaye Rambojun*: Segmentation of Psoriatic Arthritis (PSA) x-rays.
10. Benjamin Robinson: Stochastic optimal control problems related to martingale optimal transport.
11. Cameron Smith: The auxiliary region method: solving second-order reaction systems.
12. Sebastian Stolze: Bayesian networks for quantitative risk analysis.
13. Matthew Thomas*: Data integration for high-resolution, continental-scale estimation of air pollution concentrations.
14. Hayley Wragg: The ray-launching method applied to ultra-high frequency electromagnetic wave propagation.
The University of Bath is situated in the World Heritage city of Bath in the South West of England. The University overlooks the city, and has easy transport links with the city centre.
By train: Bath Spa station is situated in the city centre. There are frequent services from London, Bristol and the south coast. From here, the U1 bus can be picked up from outside of Sainsburys on the opposite side of the road from the main entrance of the station. Alternatively, the university is a walk of around 35-40 minutes away.
By coach: National express and Megabus serve Bath. See their respective websites for more information. National express coaches stop at the bus station, which is next door to the train station. Onward travel directions are then the same as for the train. Megabus coaches stop on campus.
By car: Bath is situated just off the M4. There is limited parking at the university. Please see the university's guide to parking on campus for more information.
The majority of the conference will take place in the 4 West building (Mathematical Sciences) which can be found on the parade. Registration on Tuesday morning will take place in the Chancellor's Building, which is beside the main bus stop. Please see the map below for these locations.
We invite you to the SAMBa Summer Conference in a spirit of curiosity, friendliness, open-mindedness and respect. We value your participation and want to ensure a welcoming and safe environment for all. In line with University of Bath policy, we will not tolerate harassment in any form. All participants at our event are required to agree to the following code of conduct, which can be found here.
The organisers would like to thank our contributors who helped to make this event possible: