Statistics, Probability, Analysis and Applied Mathematics (SPAAM)
SPAAM Seminar Series 2025/26
The Statistics, Probability, Analysis and Applied Mathematics (SPAAM) seminar series will take place on Thursdays between 3-4pm in room B3.02 and virtually on the SPAAM Microsoft Teams ChannelLink opens in a new window during term time. It will host a variety of talks from PhD students involved in applied mathematics research at Warwick and invited guests from other institutions (see below for the schedule and talk abstracts).
The seminars will usually host two speakers (unless otherwise stated) with each talk taking around 15-20 minutes with 5-10 minutes of questions afterwards. Speakers and committee members will hang around after the talks for social tea/coffee (and cakes!) and further questions.
This seminar series is hosted by the Warwick 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 YouTube channelLink opens in a new window to find the recorded talks!
If you would like to give a talk this academic year, please contact one of our seminar organisers:
Sasha Glendinning ()
or
Zeynep Tunalioglu ()
and we will find you a slot!
Find out more about the
Warwick SIAM-IMA Student Chapter
| Date |
Talk 1 |
Talk 2 |
|---|---|---|
| 16th October (Week 2) | Eva Zaat | Ruibo Kou |
| 23rd October 2025 (Week 3) | Chess Tournament Social | |
| 30th October 2025 (Week 4) |
Jiayao Shao |
Matthew Adeoye |
| 6th Nov 2025 (Week 5) |
Freddie Jensen |
Marc Truter |
| 13th November 2025 (Week 6) |
Ziyang Liu |
Usman Ladan |
| 20th November 2025 (Week 7) |
Hefin Lambley |
|
| 27th November 2025 (Week 8) |
Andrew Nugent |
Ramon Nartallo-Kaluarachchi |
| 4th December 2025 (Week 9) |
Mario Kart Social | |
| 11th December 2025 (Week 10) | Grega Saksida |
Shreya Sinha Roy |
| 22nd January 2026 (Week 2) | Tarek Acila | Sam Turley |
| 29th January 2026 (Week 3) | Mario Kart Social | |
| 5th February 2026 (Week 4) | Elliot Vincent | Rob Sunnucks |
| 12th February 2026 (Week 5) | Olayinka Ajayi | Daniel Higgins |
| 19th February 2026 (Week 6) | ||
| 26th February 2026 (Week 7) | ||
| 5th March 2026 (Week 8) | ||
| 12th March 2026 (Week 9) | ||
| 19th March 2026 (Week 10) | ||
Term 2 Abstracts
Week 5.
Olayinka Ajayi (University of Sussex) | Rotational-Invariant Graph Neural Network for Skeleton-based Human Activity Recognition
Human activity recognition (HAR) is crucial for applications spanning security, healthcare, and human-computer interaction. Traditional methods rely on 3D skeletal data but face challenges with viewpoint variability, as the same action can appear differently from different angles. Current graph neural networks (GNNs) in HAR often lack true rotation-invariance, necessitating extensive data augmentation and retraining.
My research introduces a novel approach by developing a rotation-invariant graph neural network for skeleton-based HAR. By leveraging learnable explicit geometric transformations such as SO(2) rotations, scaling and shearing into learned spatial embeddings. The proposed method ensures that skeletal representations remain consistent regardless of viewpoint. This innovation not only enhances recognition accuracy but also reduces the need for large, multi-angle datasets, and the additional transformations make it suitable for “in-the-wild” datasets like Kinetics400.
Daniel Higgins | Optimal vaccination strategy in a heterogeneous population
Standard epidemiological models make significant simplifying assumptions, one of which is that individuals in a population are assumed to be homogeneous. I will be discussing a model which relaxes this assumption, giving rise to a system of partial differential equations in the mean-field limit. Based on this system, I will then discuss optimal vaccine rollout, and how the results we have found may be applied in the real-world.
Week 4.
Elliot Vincent (MathSys) | Towards a pesticide-free world: behavioural modelling of sustainable crop management practices
At present, pesticides are relied upon for the prevention and control of crop diseases globally. Without robust management these diseases pose a severe threat to food supplies. However, pesticides themselves are highly toxic to the ecosystem; and furthermore, their overuse is driving the widespread development of pesticide-resistant disease strains. As a result, much global effort is being put towards finding alternative, more sustainable ways to manage crop diseases.
Crucially, while a major consideration is ensuring these methods can effectively control disease, another important consideration is whether farmers, driven by economic constraints, would actually be motivated to start using these methods. In my PhD, I use mathematical modelling to examine the adoption of one such sustainable practice: Integrated Pest Management (IPM). In this talk I'll give an overview of my behavioural model of farmer responses to profit, and the outcomes give by this model. I'll then compare my model predictions of IPM adoption with policy targets.
Rob Sunnucks (MathSys) | Exploring population models of tsetse flies to inform vector control for Human African Trypanosomiasis
Human African Trypanosomiasis (HAT) is a neglected tropical disease spread by the tsetse fly. Vector control is used to kill tsetse to lower HAT transmission, and the effectiveness of this intervention depends on assumptions about tsetse population dynamics. In this talk, I will introduce various tsetse models and explore their strengths and weaknesses, focusing on how accurately they can describe the on the ground situation, and can be fit to existing data.
Week 2.
Tarek Acila (Warwick Mathematics Institute) | Nonequilibrium as a Driver of Complex Patterned Steady States
Many physical and biological systems are often well-approximated as being in thermodynamic equilibrium, which provides a useful framework for understanding simple pattern formation and energy minimisation. However, some systems display complex spatial patterns that cannot be explained by equilibrium alone. Nonequilibrium processes can drive a system away from equilibrium and stabilise structures of finite size. In this talk, microdomains of cell membranes, known as lipid rafts, are used as an illustrative example to explore how active processes control spatial organisation. Reaction–diffusion and phase-separation models, supplemented with numerical simulations, are used to investigate how simple physical mechanisms can generate complex patterned steady states.
Sam Turley (MathSys) | Causal Entropic Force - Intelligent Decision Making
I will be introducing Causal Entropic Forces, which introduced intelligent decision-making by driving systems to maximize the diversity of possible future states. We will look at toy examples and model systems before moving onto the application to collective motion. Here, each particle will move and re-orient following this intelligent decision-making framework, and we'll observe some of the complex emergent dynamics that arise.
Term 1 Abstracts
Week 2.
Eva Zaat (Warwick Mathematics Institute) | Making Maths 'Easier'
Maths isn't always easy, but sometimes it is harder than it needs to be. In this talk we will discuss some practical strategies to make maths communication and problem solving more accessible for everyone. We will start with how mathematicians solve their problems, before exploring how maths anxiety, neurodivergence and (other) disabilities can impact how we engage with maths. By understanding these different experiences better, we can support ourselves, our collaborators, our audiences and our students; reduce barriers to learning maths and create more inclusive maths spaces.
Ruibo Kou (Warwick Mathematics Institute) | The Stochastic Casimir Effect
We model the one-dimensional ‘classical’ vacuum by a system of annihilating Brownian motions on R with pairwise immigration. A pair of reflecting or absorbing walls placed in such a vacuum at separation L experiences an attractive force which decays exponentially with L. This phenomenon can be regarded as a purely classical Casimir effect for a system of interacting Brownian motions.
Week 4.
Jiayao Shao (Warwick Mathematics Institute) | Stochastic Dynamical System Methods applied to ship capsize problem
Week 5.
Freddie Jensen (Warwick Mathematics Institute) | Observations from modelling nonlinear acoustics in brass instruments
We discuss interesting features of a recently-developed model which combines weak nonlinearity and complex geometry in duct acoustics without flow, with applications to sound in brass instruments. Topics discussed here include curvature-induced plane-wave tunnelling, a method of quantifying the speed of sound around bends, an ambiguity around forward/backward decomposition, and a new test case in the study of nodes and turning points.
Marc Truter (Warwick Mathematics Institute) | Using Deep Reinforcement Learning to Build a Periodic Table of Shapes
In this talk we show how deep reinforcement learning can be used to overcome the computational challenges faced by conventional search algorithms in the discovery of new Fano hypersurfaces. Varieties are the core objects of study in algebraic geometry, they are the geometric shapes defined by the solutions to polynomial equations. An ‘atomic’ and important class of these are called Fano varieties. Periodic tables of them are known in dimensions 1 and 2, and partially known in dimension 3. The goal of my project is to help build a periodic table in dimension 4.
Week 6.
Ziyang Liu (Warwick Mathematics Institute) | An Ito’s formula of the stochastic heat flow
The 2d critical stochastic heat flow is the solution to the multiplicative stochastic heat equation in the critical dimension 2. In a recent work of Makoto Nakashima, the stochastic heat flow is considered in a martingale setting, hence leading to a study of an associated Ito’s formula. In this talk, I will briefly introduce the stochastic flow as a scaling limit of a discrete polymer model and discuss its martingale structure and associated Ito’s formula.
Usman Ladan (Warwick Department of Statistics) | An introduction to branching processes and their approximations
In this talk, we introduce branching processes and discuss the behaviour of both discrete and continuous time models. These objects have applications in a wide variety of areas such as biology and physics. We will also discuss how to efficiently approximate these objects via mutation/selection type schemes.
Week 7.
Hefin Lambley (Warwick Mathematics Institute) | Learning in function space: theory and practice
I will discuss neural operators: generalisations of neural networks to inputs and outputs that are functions rather than vectors. I will show that many existing machine-learning methods can be lifted to function space, and explain some challenges in doing so. I will then show some of the highly successful applications of neural operators in scientific machine learning to regression and generative modelling.
Sam Turley (Warwick Mathematics Institute) | Causal Entropic Force - Intelligent Decision Making
I will be introducing Causal Entropic Forces, which introduced intelligent decision-making by driving systems to maximize the diversity of possible future states. We will look at toy examples and model systems before moving onto the application to collective motion. Here, each particle will move and re-orient following this intelligent decision-making framework, and we'll observe some of the complex emergent dynamics that arise.
Week 8.
Andrew Nugent (UCL) | Existence of stationary distributions in an age-structured model of opinion formation
Models of opinion formation describe how a population of agents interact and update their opinions on some topic. I will first introduce an SDE model for a population with age structure, in which agents die and are replaced by new agents with random initial opinions. While this model displays the clustering that is typical in opinion dynamics, it is interesting that these clusters persist far beyond the lifetimes of individual agents. Such macroscopic patterns correspond to stationary distributions of the mean field PDE, arising in the large population limit. This talk will explain our approach to rigorously establishing the existence and properties of these stationary distributions.
Week 9.
Grega Saksida (Warwick Mathematics Institute) | Graphic representations of spin systems
Physicists have come up with many models to explain how magnetic properties of materials are linked to their microscopic structure. Common to all these models is the description of materials as particles arranged in a lattice. Every particle possesses a quantum-mechanical quantity called spin, and we call a collection of such particles a spin system. Particles with spin generate a magnetic field, and when enough particles' spins are aligned, the magnetic field becomes macroscopic. We then say the system is ordered.
I will present a few examples of spin systems, and show how we define the notion of order. I will then sketch how one can study the latter by counting graphs.
Shreya Sinha Roy (Warwick Department of Statistics) | Generalized Bayesian deep reinforcement learning
Bayesian reinforcement learning (BRL) is a method that merges principles from Bayesian statistics and reinforcement learning to make optimal decisions in uncertain environments. As a model-based RL method, it has two key components: (1) inferring the posterior distribution of the model for the data-generating process (DGP) and (2) policy learning using the learned posterior. We propose to model the dynamics of the unknown environment through deep generative models, assuming Markov dependence. In the absence of likelihood functions for these models, we train them by learning a generalized predictive-sequential (or prequential) scoring rule (SR) posterior. We used sequential Monte Carlo (SMC) samplers to draw samples from this generalized Bayesian posterior distribution. In conjunction, to achieve scalability in the high-dimensional parameter space of the neural networks, we use the gradient-based Markov kernels within SMC. To justify the use of the prequential scoring rule posterior, we prove a Bernstein-von Mises-type theorem. For policy learning, we propose expected Thompson sampling (ETS) to learn the optimal policy by maximising the expected value function with respect to the posterior distribution. This improves upon traditional Thompson sampling (TS) and its extensions, which utilize only one sample drawn from the posterior distribution. This improvement is studied both theoretically and using simulation studies, assuming a discrete action space. Finally, we successfully extended our setup for a challenging problem with a continuous action space without theoretical guarantees.