Junior Applied Mathematics Seminar (JAMS)
JAMS Series 2026/27 (Previously SPAAM)
The Junior Applied Mathematics Seminar (JAMS) series will take place every other Wednesday between 4-5pm in room B3.03 during term time. The seminar can also be joined online . It will host a variety of talks from PhD students involved in applied mathematics research at 91福利 and invited guests from other institutions (see below for the schedule and talk abstracts).
If you are interested in speaking, please use the !
The seminars will usually host three speakers (unless otherwise stated) with each talk taking around 15 minutes with 5 minutes of questions afterwards. Speakers and committee members will hang around after the talks for a social visit to Varsity pub and further questions.
This seminar series is hosted by the 91福利 SIAM-IMA Student ChapterLink opens in a new window. Please do contact one of the committee members if you would like to join and be added to the MS Teams channel. Note that these talks may be recorded so do join with audio and video off if you don't wish to feature!
If you missed the seminar, head over to our to find the recorded talks!
If you would like to give a talk this academic year, please , or contact our seminar organiser:
Rob Sunnucks (rob.sunnucks@warwick.ac.uk)
Find out more about the
91福利 SIAM-IMA Student Chapter
| Date |
Talk 1 |
Talk 2 |
Talk 3 |
|---|---|---|---|
| 7th October (Week 1) | Luca Chadwick | Jimmy Kornelije Gunnarsson |
Gracie Allen |
| 21st October (Week 3) | Benedict Russell | Joshua Looker | Rob Sunnucks |
| 4th November (Week 5) | Alastair Crossley | Peter Matthews | Vinayak Niraj |
| 18th November (Week 7) | TBD | TBD | TBD |
Term 1 Abstracts
Week 5
Alastair Crossley | Statistics | A Stochastic Approach to Infinite-Rate Electron鈥揚hoton Cascades
Electron鈥損hoton cascades arise when high-energy electrons and photons interact through bremsstrahlung and pair production, generating branching populations of secondary particles. While classical cascade equations describe the evolution of expected particle populations, we instead begin with the underlying stochastic dynamics.
In this talk, I will introduce a measure-valued branching process describing electron鈥損hoton cascades. The main mathematical challenge is the infinite rate of bremsstrahlung emissions, which requires additional care in constructing the process. Using energy estimates, we establish well-posedness and derive a semimartingale representation. I will then briefly explain how the first-moment generator connects the stochastic model to classical electron鈥損hoton cascade equations.
Peter Matthews | Statistics | The Proton Therapy Model Zoo
We will discuss the various models used for different purposes within proton therapy, along with their mathematical foundations. We will also talk about why we need to use a variety of different models for different tasks.
Vinayak Niraj | Statistics | Tempered Bouncy Particle Sampler for Heavy Tailed sampling
Piecewise Deterministic Markov Processes (PDMPs) are stochastic processes initially proposed in physics and chemistry and later applied to MCMC (Markov Chain Monte Carlo) and ML. Most standard MCMC methods fail at sampling heavy-tailed target distributions. We generalise a standard PDMP (the Bouncy Particle Sampler) to optimally sample heavy-tailed targets and prove fast convergence to the target distribution.
Week 3
Benedict Russell | Dynamics of a learning population with partner selection
We develop an analytical model of policy-gradient learning with partner selection, showing how choosing interaction partners changes incentives and promotes cooperation. We prove that population variance is necessary for cooperation to emerge, and use stochastic differential equations to derive conditions favouring cooperation and establish the existence of a stationary distribution. Simulations validate the model and reveal how learning rates influence cooperative behaviour.
Joshua Looker | Developing early warning signals for spatial epidemic models
There is a growing field of research showing the ability of early warning signals (EWS) to anticipate transitions in epidemic models. However, consideration of the impacts of spatial heterogeneity on the signals is minimal.
In this talk we will introduce a basic spatial interacting particle system and derive the PDE and SPDE corresponding to the deterministic large population limit and fluctuations around it. We will derive the EWS and show through simulations how they differ to the well-mixed case.
Rob Sunnucks | Bayesian decision making for vector control of sleeping sickness
Gambiense human African trypanosomiasis, otherwise known as sleeping sickness, is a disease in West and Central Africa spread by tsetse. A key method to combat the spread of the disease is the deployment of tiny targets which kill tsetse. Monitoring of the tsetse population can allow us to ascertain the effectiveness of tiny targets in a given area and use this information to inform whether deployments should continue. However, the usefulness of this monitoring has been called into question.
By combining model projections with Bayesian updating of parameter distributions, we shall show how to calculate the monetary benefit of this monitoring in reducing the burden of sleeping sickness and lowering costs, utilising the active adaptive management decision-making framework. In doing so, we shall demonstrate when and where this data-driven decision-making can prove cost-effective.
Week 1
Gracie Allen | Statistics | The Sound Underground: Learning 91福利 Earthworms with Acoustics鈥�
Earthworms are widely considered beneficial ecosystem engineers, but their population trends and influences remain uncertain due to limited historical monitoring data and problems with standard methods. Soil acoustic monitoring is a recently developed non-invasive approach which may address this, but a statistically sound framework for inferring earthworm behaviour and abundance from acoustic recordings remains underdeveloped. I investigate how soil acoustic recordings can be used to build such a framework, using controlled laboratory recordings of known earthworm densities. Signal de-noising and exploratory clustering are used to detect and classify earthworm signals, while discrete state stochastic processes provide a framework to model earthworm activity in both individual and multiple-worm cases. This work establishes a foundation for developing robust acoustic methods for monitoring earthworm populations and assessing changes in soil fauna, which will be scaled beyond the lab and into the field.
Luca Chadwick | Mathematics | Empirical Predictively Oriented Posteriors
Bayesian inference is often designed for parameter estimation, but many applications prioritise predictive performance. Predictively Oriented (PrO) posteriors address this by directly optimising predictive distributions under scoring rules and can offer improved robustness under model misspecification. We introduce a particle-based optimisation approach for computing PrO posteriors through joint gradient descent over posterior distributions and hierarchical hyperparameters. Using a Wasserstein鈥揈uclidean geometry, we derive an interacting particle system that approximates the corresponding gradient flow. We establish theoretical properties including existence and uniqueness of minimisers.
Jimmy Kornelije Gunnarsson | Lund University, Mathematics | Discontinuous Galerkin Weighted Interior Penalty Method with a Provable Discrete Maximum Principle for the Cahn鈥擧illiard Equations for arbitrary polynomial order
Symmetric Weighted Interior Penalty (SWIP) discretization, was introduced by (Ern et al., 2009) and applied to the Cahn鈥擧illiard equations in (Gunnarsson & Kl枚fkorn, 2026a), where a specific weight option was proposed and compared to other schemes which numerically provide evidence of structure preservation. For the newly developed scheme in (Gunnarsson & Kl枚fkorn, 2026c) a discrete maximum principle and mass conservation is proven, and energy dissipation is observed. In this talk we present recent findings both on the theoretical level and on the numerical level. We further provide some generalizations which follow from coupling via advection -- a similar context was studied in the hp-adaptive case in (Gunnarsson & Kl枚fkorn, 2026b).