The SIDur Exec are pleased to announce the programme for the SIDur PGR Symposium on Tuesday February 17th from 10:30 in MCS 1022 (VisLab) in the Maths & Computer Science Building.
Minor change: Patrick Creagh’s talk has now been moved to 15:00 and Andreo Chimal’s talk brought forward to 11:00.
Programme
10:30 – 11:00 Welcome & Coffee
11:00 – 11:20 Wind farm wake modelling:from linearising equations to avoiding ‘wind theft’
Andreo Chimal, Durham Maths, Applied Mathematics
Abstract: Wind is a strategic source of energy for countries with no reliable oil and gas reserves, such as the UK and most of the EU. It is clean and sustainable, produces no atmospheric emissions, and does not pollute water during operation. Its main drawback is that wind energy production can be inefficient, with one of the main factors being the wake effect, also known as “wind theft”: wind is slowed down as it passes through a wind turbine, so it is slower behind a wind farm as a whole than in front of it. The reduced wind speed, combined with the fact that energy is transferred to smaller-scale eddies, disrupts the airflow through downstream wind farms and reduces the amount of energy that a second wind farm can recover. Our research group has developed a model that couples the atmospheric boundary layer with free-atmosphere wave dynamics through the capping inversion layer, which I will present. From this model, we obtained very efficient 2D and 3D numerical schemes that simulate the wake created downstream of an object in the boundary layer (not limited to wind farms, we can also study individual turbines or even cities), the results of which I will also show.
Slides here
11:20 – 11:40 A ferociously brief history of optimal transport
Guy Parker, Durham Maths, GL_n(\R)
Abstract: This is a ferociously brief history of optimal transport.
11:40 – 12:00 Coffee Break
12:00 – 12:20 From Sums of Squares to Theta Series
Samuel Shepherd, Durham Maths, Algebra and Number Theory
Abstract: We discuss two well-known classical theorems in elementary number theory dating to the 17th and 18th centuries, and illustrate how these lead naturally to a first encounter with theta series, an important class of examples of `modular forms’, holomorphic functions on the complex upper-half plane satisfying a certain transformation condition.
Slides here
12:20 – 12:40 Gaussian Processes for Probabilistic Seismic Wavefield Simulation
Oscar Humphreys, Durham Earth Sciences, Geophysics
Abstract: We introduce a novel, scalable Gaussian process framework for simulating general seismic wavefield observables given limited information from near source receivers. The generality of Gaussian process imbues it with strong links to other machine learning and regression methods, and can be leveraged to derive information on the source parameters implicit in the training data, as well as producing an efficient encoding of the data. The framework can be applied to multivariate functions efficiently, and represent weakly non-Gaussian posteriors. The framework uses deep kernel learning and inducing point variational inference to scale it to high dimensional data.
Slides here
12:40 – 14:00 Lunch
14:00 – 14:20 Convex Hulls of Stochastic Processes
Charlie Hall, Durham Maths, Probability
Abstract: The convex hull of a non-empty set $A\subset\mathbb{R}^n $ is the smallest compact, convex set containing $A$. In this talk, we study the evolution of the convex hull of a random set over discrete or continuous time. Our goal is to establish almost sure results concerning convergence in the Hausdorff metric, as well as set containment properties of the resulting random convex sets.
Slides here
14:20 – 14:40 Heat transport in rotating, sheared convection
Paraskevi Katsiavria, Durham Maths, Applied Mathematics
Abstract: Convection drives heat transport in stars and giant planets, yet its behaviour under the combined effects of rotation and shear is not well understood. Existing models fail when shear flows are strong. Using two-dimensional nonlinear simulations, we investigate how shear modifies rotating convection and its heat transport.
Slides here
14:40 – 15:00 Coffee Break
15:00 – 15:20 An Introduction to the Langlands Program
Patrick Creagh, Durham Maths, Pure/Arithmetic Study Group
Abstract: The Langlands Program is a large web of conjectures spanning the intersection of modern algebra, number theory and analysis. Perhaps the most famous result to come out of the program is Wiles’s proof of Fermat’s last theorem, achieved by somehow relating elliptic curves to modular forms in a meaningful way. In this talk, I hope to introduce an example of one of these correspondences and provide an insight into the motivation and philosophy at play in modern day research of one of the aspects of the program.
Slides here
15:20 – 15:40 Understanding GPUs by Forgetting They’re Special
Pawel Radtke, Durham Computer Science, Scientific Computing
Abstract: GPUs are often presented as exotic, opaque devices that require entirely new ways of thinking about performance. This talk takes the opposite approach. By temporarily “forgetting” what makes GPUs special, much of the performance intuition already developed for CPUs can be reused: how work is scheduled, how data moves, and why locality and parallelism matter.
The talk uses familiar CPU concepts—such as cores, threads, SIMD, memory hierarchy, and throughput—to introduce their GPU counterparts, showing that many performance pitfalls disappear once the hardware is viewed through a shared mental model. Rather than focusing on concrete APIs or optimisation recipes, the emphasis is on making GPU behaviour familiar and predictable, rather than discovered through trial and error.
Slides here
15:40 – 16:00 Closing Remarks & Group Photo