Seminar Series Logo

AY 2025-2026 Semester 1 Seminar Schedule

Timetable

Date

Day

Time

Week no.

Presenter

Affil.

Dept. Contact

Title

2025-09-11

Thursday

11am

1

Hlib Husarov & Eberhard Mayerhofer

Mary Immaculate Secondary School, Lisdoonvarna & University of Limerick

Eberhard Mayerhofer

An algorithm for the product-sum equality

2025-09-17

Wednesday

3pm

2

Márton Karsai

Central European University

Padraig McCarron

Behaviour-driven epidemic phenomena in heterogeneous populations and networks

2025-09-26

Friday

3pm

3

Nico Gray

University of Manchester

Andrew Fowler

Coupling rheology and segregation in granular flows

2025-10-09

Thursday

11am

5

Lisa McFetridge

Queen's University Belfast

Shirin Moghaddam

Advances in Robust Mixed Effects Modelling

2025-10-16

Thursday

3pm

6

Dingjia Cao

University of Limerick

James Sweeney

A wavelet lifting approach for spatial/network data

2025-10-23

Thursday

11am

7

Theresa Smith

University of Bath

James Sweeney

Multi-output gaussian process models for pollution in river networks

2025-11-06

Thursday

11am

9

Yanghong Huang

University of Manchester

Mehakpreet Singh

Finite Difference Methods for fractional Laplacian of radial functions

2025-11-13

Thursday

11am

10

Lea Kaufmann

RWTH Aachen University

Kevin Burke

Factor Selection and Levels Fusion in High-Dimensional Logistic Regression

2025-11-19

Wednesday

12:30pm

11

Shresh Jain

University of Manchester

Doireann O'Kiely

Investigating Material Nonlinearities in Coke Cans and Plastic Holey Columns

2025-11-20

Thursday

11am

11

Alison O'Connor

University of Limerick

David O'Sullivan

Hip Fractures Trends in Older Adults in Ireland

2025-11-25

Tuesday

2pm

12

Christos Xenophontos

University of Cyprus

Natalia Kopteva

Neural Networks for singular perturbations

2025-11-27

Thursday

3pm

12

Hugo Luiz Oliveira

University of Campinas (Unicamp), Brazil

Michael Vynnycky

Numerical and computational modelling of the Wheatley aortic valve

2025-12-04

Thursday

11am

13

Silvia D'Angelo

Trinity College Dublin

David O'Sullivan

Clustering high-dimensional discrete data: a compressed approach

Abstracts

As seminar abstracts become available from the speakers this page will be updated.


Seminar week 1 by Hlib Husarov & Eberhard Mayerhofer

Date: 2025-09-11 at 11am
Speaker: Hlib Husarov & Eberhard Mayerhofer (Mary Immaculate Secondary School, Lisdoonvarna & University of Limerick)
Host: Eberhard Mayerhofer

Title: An algorithm for the product-sum equality

Abstract: We propose a recursive algorithm for identifying all finite sequences of positive integers whose product equals their sum. Our method uses solutions of strictly shorter length that are iteratively extended in pursuit of a valid solution. The algorithm is efficient, with a time complexity similar to the quick sort algorithm. Arxiv: https://arxiv.org/abs/2508.09647

Speaker’s Google Scholar

Speaker’s Paper


Seminar week 2 by Márton Karsai

Date: 2025-09-17 at 3pm
Speaker: Márton Karsai (Central European University)
Host: Padraig McCarron

Title: Behaviour-driven epidemic phenomena in heterogeneous populations and networks

Abstract: While realistic approaches have become increasingly important in epidemic modelling, behavioural factors and individual differences have historically been overlooked due to the lack of high-resolution data and appropriate mathematical methods. This gap became particularly evident during the recent pandemic, highlighting the need for large-scale data collection on individual-level epidemic-related behaviours across representative populations. These advancements have revealed several new and interesting spreading phenomena that challenge our previous understandings. In this talk, we will focus on a few examples of such new insights, derived from data-driven and behaviour-informed epidemic models. We will explore the effects of seeding strategies and input-output inequalities in spreading dynamics, and the paradoxical effects of awareness-driven adaptive behaviour on epidemic outcomes. These findings highlight the importance of data-driven models incorporating behavioural factors, offering a more accurate understanding and modelling of real-world epidemic scenarios.

Speaker’s Google Scholar

Speaker’s Paper 1 Speaker’s Paper 2 Speaker’s Paper 3


Seminar week 3 by Nico Gray

Date: 2025-09-26 at 3pm
Speaker: Nico Gray (University of Manchester)
Host: Andrew Fowler

Title: Coupling rheology and segregation in granular flows

Abstract: During the last fifteen years there has been a paradigm shift in the continuum modelling of granular materials; most notably with the development of rheological models, such as the μ(I)-rheology (where μ is the friction and I is the inertial number), but also with significant advances in theories for particle segregation. This paper details theoretical and numerical frameworks (based on OpenFOAM) which unify these currently disconnected endeavours. Coupling the segregation with the flow, and vice versa, is not only vital for a complete theory of granular materials, but is also beneficial for developing numerical methods to handle evolving free surfaces. This general approach is based on the partially regularized incompressible-rheology, which is coupled to the gravity-driven segregation theory of Gray & Ancey (J. Fluid Mech., vol. 678, 2011, pp. 353-588). These advection-diffusion-segregation equations describe the evolving concentrations of the constituents, which then couple back to the variable viscosity in the incompressible Navier-Stokes equations. A novel feature of this approach is that any number of differently sized phases may be included, which may have disparate frictional properties. Further inclusion of an excess air phase, which segregates away from the granular material, then allows the complex evolution of the free surface to be captured simultaneously. Three primary coupling mechanisms are identified: (i) advection of the particle concentrations by the bulk velocity, (ii) feedback of the particle-size and/or frictional properties on the bulk flow field and (iii) influence of the shear rate, pressure, gravity, particle size and particle-size ratio on the locally evolving segregation and diffusion rates. The numerical method is extensively tested in one-way coupled computations, before the fully coupled model is compared with the discrete element method simulations of Tripathi & Khakhar (Phys. Fluids, vol. 23, 2011, 113302) and used to compute the petal-like segregation pattern that spontaneously develops in a square and triangular rotating drums.

Particle-size segregation patterns in a partially filled triangular rotating drum. E.S.F. Maguire, T. Barker, M. Rauter, C.G. Johnson & J.M.N.T. Gray (2024) J. Fluid Mech 979, A40. (movie 1) (movie 2) (movie 3) (movie 4) (movie 5) (movie 6) (movie 7) (movie 8) (movie 9) (movie 10) (movie 11) (movie 12) (movie 13) (movie 14) (movie 15) (movie 16) (movie 17) (movie 18) (movie 19)

Coupling rheology and segregation in granular flows T. Barker, M. Rauter, E. S. F. Maguire, C. G. Johnson & J. M. N. T. Gray (2021) J. Fluid Mech. 909, A22. (movie 1) (movie 2) (movie 3) (movie 4) (movie 5) (movie 6)

An experimental scaling law for particle-size segregation in dense granular flows T. Trewhela, C. Ancey & J. M. N. T. Gray (2021) J. Fluid Mech. 916, A55. (movie 1) (movie 2)

Particle Segregation in Dense Granular Flows J. M. N. T. Gray (2018) Ann. Rev. Fluid Mech. 50, 407–433. (movies)

Partial regularisation of the incompressible μ(I)-rheology for granular flow T. Barker & J. M. N. T. Gray (2017) J. Fluid Mech. 828, 5–32. (movies)


Seminar week 5 by Lisa McFetridge

Date: 2025-10-09 at 11am
Speaker: Lisa McFetridge (Queen’s University Belfast)
Host: Shirin Moghaddam

Title: Advances in Robust Mixed Effects Modelling

Abstract: Over the past few decades, rapid technological innovation has revolutionised data collection, in particular in medicine where wearable devices are now regularly used to continuously track individual health metrics. This explosion of longitudinal data presents both opportunities and challenges for statistical modelling.

Mixed effects models have become a cornerstone for analysing repeated measurements over time. Yet, their widespread use rests on a critical assumption – that random effects and error terms follow a normal distribution. Previous research has shown that when longitudinal outliers are present, these assumptions can lead to biased estimates and imprecise variance components, undermining the reliability of clinical insights.

To address this, robust modelling techniques have emerged that replace normality assumptions with t-distributed random terms. These models adaptively estimate degrees of freedom from the data itself, automatically reverting to normality when appropriate, providing flexibility without loss of precision or increase in bias. The utility of such models has been demonstrated across diverse applications, from predicting ICU patient risk, to cardiac monitoring and anomaly detection in elite athletes. Crucially, individuals flagged as longitudinal outliers often exhibit significantly poorer survival or risk profiles, underscoring the clinical importance of identifying and modelling such cases accurately.

In this talk, we’ll explore cutting-edge advancements in robust mixed effects modelling, showcasing their potential in health monitoring and precision medicine.

Speaker’s Google Scholar


Seminar week 6 by Dingjia Cao

Date: 2025-10-16 at 3pm
Speaker: Dingjia Cao (University of Limerick)
Host: James Sweeney

Title: A wavelet lifting approach for spatial/network data

Abstract: As an emerging research area, analysing functions that arise from network (graph) structures attracts researchers in both statistics and signal processing communities. Data collected over graphs can be modelled as noisy observations of an unknown function over the vertices of a graph structure, fully described by its vertices and their connections, the edges.

While current literature is rich when data are collected from the graph vertex space, the data collected from graph edges call for new techniques, and in its turn reaches across many application fields, from traffic networks to neuroscience and hydrology. Wavelets can be powerful tools for understanding the behaviour of the underlying (edge) functions due to their computational efficiency and robust performance in the presence of discontinuities.

In this talk, we will begin by tracing the ‘evolution’ of wavelet methods, from classical wavelets to second-generation wavelets (constructed via the lifting scheme), and then to a variant that allows us to handle a broader class of data (such as spatial data). We will then move on to construct some new wavelet bases that can represent and perform nonparametric regression for functions defined on both the vertex and edge sets of graphs.

Speaker’s Google Scholar


Seminar week 7 by Theresa Smith

Date: 2025-10-23 at 11am
Speaker: Theresa Smith (University of Bath)
Host: James Sweeney

Title: Multi-output gaussian process models for pollution in river networks

Abstract: The impact of human activity on the quality of surface waters, particularly rivers, has recently received considerable attention. Statistical models aimed at characterizing the spatiotemporal distribution of biological and chemical indicators across river networks must address several unique challenges not typically encountered in standard spatial modelling. In this seminar, I will present a class of spatiotemporal models designed for multivariate data sampled on river networks. These models build upon the framework introduced by Ver Hoef et al. (2006), which replaces traditional Euclidean distances with stream distance metrics to better represent spatial relationships in river systems. We extend this approach to incorporate temporal dynamics and multivariate dependencies, while also addressing practical aspects of water quality monitoring such as censoring limits and intermittent sampling. Finally, I will introduce a variational inference framework for fitting these models efficiently and share results from simulation experiments comparing our proposed approach with more conventional Gaussian Process models.

Speaker’s Google Scholar


Seminar week 9 by Yanghong Huang

Date: 2025-11-06 at 11am
Speaker: Yanghong Huang (University of Manchester)
Host: Mehakpreet Singh

Title: Finite Difference Methods for fractional Laplacian of radial functions

Abstract: Numerical evaluation of nonlocal operators like the fractional Laplacian is more computationally intensive because of the dependence on the underlying function over the whole space. On the other hand, many solutions to the fractional counterparts of classical semi-linear PDEs, especially ground states obtained via variational methods, are radial. In this talk, fractional Laplacian of radial functions in general dimensions will be considered with a kernel represented by a Gauss hypergeometric function of the radial variables. The singular part of the kernel is isolated and then treated with effective methods well studied in the one-dimensional context, while the regular part can be evaluated with by classical quadrature. The method can be extended to general non-radial functions that can be expanded using spherical harmonics, making it effective in the numerical study of fractional equations and complementing existing theoretical investigations.

Speaker’s Google Scholar


Seminar week 10 by Lea Kaufmann

Date: 2025-11-13 at 11am
Speaker: Lea Kaufmann (RWTH Aachen University)
Host: Kevin Burke

Title: Factor Selection and Levels Fusion in High-Dimensional Logistic Regression

Abstract: In the last decades, high-dimensional problems in a regression-type context including a large number of explanatory variables arise in a huge variety of application fields. Specifically considering categorical explanatory variables, i.e., factors, the dummy-coding scheme introduces one (dummy) variable for each factor level, thus the dimension of the parameter space grows rapidly, even for a moderate number of factors. To simultaneously obtain the estimates of the regression coefficients and reducing the dimension of the parameter space, penalized regression is known to be a suitable tool. In this talk, I will discuss the characteristics that a penalization technique needs to fulfil to be suitable for the application to factors, as it is not only important to select those having an influence on the response variable (factor selection), but also to fuse those levels of a factor having a similar influence (levels fusion). A new penalization technique, called L0-Fused Group Lasso (L0-FGL), is introduced, tailored for the needs of factors. The quality of this new method is underlined by its theoretical properties, e.g., root-n-consistency and asymptotic normality, as well as by its performance in simulation studies.

Speaker’s Paper

Speaker’s Webpage


Seminar week 11 by Shresh Jain

Date: 2025-11-19 at 12:30pm
Speaker: Shresh Jain (University of Manchester)
Host: Doireann O’Kiely

Title: Investigating Material Nonlinearities in Coke Cans and Plastic Holey Columns

Abstract: Most real-world buckling problems, from Coke cans to metamaterials, involve metals or other hard materials that deform plastically. Yet, many studies focus primarily on geometric nonlinearities, often neglecting the material nonlinearities that play a crucial role in determining post-buckling behaviour. In this talk, we examine two systems in which we incorporate these effects through semi-empirical modelling of material nonlinearities: hard holey columns and the sequential ring-buckling of fluid-filled Coke cans. By including material nonlinearities in our models, we show that intuitive, mechanics-based approaches can successfully capture the complex behaviour observed in the buckling of hard materials.

Speaker’s Webpage


Seminar week 11 by Alison O’Connor

Date: 2025-11-20 at 11am
Speaker: Alison O’Connor (University of Limerick)
Host: David O’Sullivan

Title: Hip Fractures Trends in Older Adults in Ireland

Abstract: European healthcare systems are under pressure from rapidly ageing populations, with hip fractures representing one of the most costly and impactful injuries among older adults (1,2). Ireland faces the dual challenge of both a rapidly growing ageing population and one of the highest European rates in hip fracture incidence (3,4). Irish standards require admission to orthopaedic care within four hours, yet national audits show this target is rarely achieved (5). Improving hip fracture services requires not just clinical reform but also new ways of leveraging data.

In this talk, we show how publicly available datasets can be used to uncover correlations and trends in hip fracture demand. When combined with national audit records we show how population health interacts with environmental pressures. Our findings suggest that extreme weather is a strong national indicator of surges in hip fractures with some regions experiencing more pressure than others. This creates opportunities not only for predictive modelling and smarter resource allocation, but also for preventative strategies where timely information can be used to warn and protect at-risk groups.

References

  1. Looi MK. The European healthcare workforce crisis: how bad is it? BMJ. 2024 Jan 19;384:q8.

  2. EUROSTAT. Ageing Europe - statistics on population developments. Eurostat; 2020. Report No.: 80393.

  3. CSO. Central Statistics Office (CSO). Central Statistics Office; 2022 [cited 2024 Sep 23]. Society Measuring Ireland’s Progress 2022 - Central Statistics Office. Available from: https://www.cso.ie/en/releasesandpublications/ep/p-mip/measuringirelandsprogress2022/society/

  4. Walsh ME, Ferris H, Coughlan T, Hurson C, Ahern E, Sorensen J, et al. Trends in hip fracture care in the Republic of Ireland from 2013 to 2018: results from the Irish Hip Fracture Database. Osteoporos Int. 2021 Apr 1;32(4):727–36.

  5. Kelly P, Horan B, Murphy T, Ahern E, Brent L, Kelly F, et al. Irish Hip Fracture Database National Report 2022 [Internet]. National Office of Clinical Audit (NOCA); 2022 p. 1–88. Available from:

https://d7g406zpx7bgk.cloudfront.net/38cf762d5b/irish_hip_fracture_database_national_report_2022_final.pdf

Speaker’s Google Scholar


Seminar week 12 by Christos Xenophontos

Date: 2025-11-25 at 2pm
Speaker: Christos Xenophontos (University of Cyprus)
Host: Natalia Kopteva

Title: Neural Networks for singular perturbations

Abstract: In this talk we will give an introduction to Neural Networks (NNs) and describe how they can be used for emulating the solution to singularly perturbed problems. In addition, we will present results on the expressivity rate bounds for solution sets of a model class of singularly perturbed, elliptic two-point boundary value problems, in Sobolev norms, on the bounded interval (-1,1). The expression rate bounds in Sobolev norms in terms of the NN size are robust, i.e. uniform with respect to the singular perturbation parameter for several classes of DNN architectures. In particular, ReLU NNs, tanh- and sigmoid-activated NNs. The latter activations can represent “exponential boundary layer solution features” explicitly, in the last hidden layer of the DNN, i.e. in a shallow subnetwork, and afford improved robust expression rate bounds in terms of the NN size.


Seminar week 12 by Hugo Luiz Oliveira

Date: 2025-11-27 at 3pm
Speaker: Hugo Luiz Oliveira (University of Campinas (Unicamp), Brazil)
Host: Michael Vynnycky

Title: Numerical and computational modelling of the Wheatley aortic valve

Abstract: Recent studies have indicated an increase in cardiovascular diseases in developing countries, particularly valvular disorders caused by rheumatic fever, which are affecting younger populations. In severe cases, when the natural valve cannot be repaired, replacing it with a prosthetic valve may be the most effective way to ensure the patient’s survival. Existing artificial valves have limitations, such as early deterioration and the risk of clot formation. To address these issues, the Wheatley Valve introduces an innovative S-shaped leaflet design that enhances the washout effect in the aortic root, thereby reducing the risk of thrombus formation and the need for extensive antiplatelet therapy. In this talk, we will discuss the main elements needed to build a fluid-structure interaction computational model able to represent the real behaviour of the Wheatley valve and which can be readily extended to other valve designs.

Speaker’s Google Scholar


Seminar week 13 by Silvia D’Angelo

Date: 2025-12-04 at 11am
Speaker: Silvia D’Angelo (Trinity College Dublin)
Host: David O’Sullivan

Title: Clustering high-dimensional discrete data: a compressed approach

Abstract: Clustering high-dimensional data is a challenging task, typically addressed in the context of continuous numerical data. Recent literature is exploring strategies to cluster high-dimensional categorical and discrete data, which finds numerous applications, such as RNA sequencing and text data analysis. We propose a fast and easy-to-implement approach to cluster high-dimensional discrete data, scalable to datasets with thousands of dimensions where other strategies may be computationally unfeasible. Our approach relies on reducing the dimension of the data by performing a deterministic compression to a drastically lower dimension. The method employs a lossy compression that reduces the data to a collection of continuous features. We demonstrate that such compressed features can be treated as approximately normally distributed, allowing the application of standard finite Gaussian mixture models for model-based clustering. We discuss the approach and study its performance on a series of simulated scenarios with different dimensions and levels of complexity, involving both categorical and count data. Additionally, we illustrate the method on real-world data.

Speaker’s Google Scholar

Speaker’s Webpage 1 Speaker’s Webpage 2