Home About Call for Posters/Talks Organisers Speaker Abstracts Posters Schedule Map

MPIE 2025 Conference

Maths Physics Informatics Engineering

Speaker Abstracts

MPIE 2025 Conference Sessions

Morning Session

Prof Michael Luck

Introduction

Niel de Beaudrap

Keynote: Quantum Computation

Quantum computation is a subject that draws together physics, as the basis of ways that matter and energy can be used to store information; mathematics, as the subject which describes the abstractions to allow us to describe how that information can be structured at a higher level; and computer science, as the basis to describe how to work with those abstractions to perform computation. One particular issue in quantum computation is the fact that quantum mechanical systems are sensitive to environmental influence, which is a challenge for controlling them well enough to compute. In this talk, I present a high-level view of one particular line of research to achieve this, from basic principles to recently published work.

I aim to provide a sense of the ways in which several mathematical concepts play a lively role in quantum information theory and quantum error correction. These include random variables, self-adjoint operators, over-categories, symplectic vector spaces, finite abelian groups, topology, homotopy, and string diagrams - in many cases presented with an accent of the finite field of order 2, and all with the motivation of figuring out how to choreograph the behaviour of tiny physical systems to compute things.

Parsa Rahimi

Hybrid Yb/Ba ion trap quantum computing with laser free coherent control

Magnetic field noise reduction for improving coherence times with a feed forward method.

Mateusz Kupper

Quantum Error Correction: String Diagrams for Defect-Based Surface Code Computing

Surface codes are a popular choice for implementing fault-tolerant quantum computing. Two-qubit gates may be realised in these codes using only nearest-neighbour interactions, either by lattice surgery or by braiding defects around each other. The effect of lattice surgery operations may be simply described using the ZX-calculus: a graphical language that has proven effective for program design and optimisation. In this work, we formalise a similar description via the ZX-calculus of defect braiding, as it is conventionally described. We define a graphical calculus KNOT, denoting the logical effects (in the absence of byproduct operations) of defect braiding in surface codes: we show how these effects may be described via a fragment of ZX-calculus which we call the (0, pi)-fragment. We then use a doubling construction to define a subtheory of KNOT, more specialised to standard encoding techniques in the defect braiding literature. Within this subtheory, we encompass standard braiding techniques by families of ribbon-like and tangle-like diagrams, each with semantics distinct from KNOT, in terms of the (0, pi)-fragment of ZX diagrams (again in the absence of byproducts). These subtheories may be used interoperably, and are each sound and complete for the (0, pi)-fragment of ZX diagrams. This provides a starting point to use the formal diagrammatics to analyse the operational effects of defect braiding procedures.

Xavier Calmet

Black Holes in the Quantum Realm

In this talk, we present recent results demonstrating that quantum physics is crucial to make sense of black holes. We explain how modern methods in quantum field theory can be used to perform calculations in quantum gravity. Our work is a first step towards merging quantum physics and Einstein's theory of general relativity.

'MPIE for Healthcare Applications' Panel

Karrie Liu

Department: Mathematics

Karrie Liu is a data scientist and healthcare analytics consultant currently undertaking a PhD in Women's Health at the University of Sussex. Her doctoral research focuses on using mathematical modelling and operational research to improve health equity through Women's Health Hubs in the UK. With over 15 years of experience in healthcare, life sciences, and data strategy, Karrie brings both technical expertise and real-world insight to the discussion. She is also a trustee at Age UK Wandsworth, where she advocates for older women's health, and she continues to lead pro bono data projects that support community-based services. Her work bridges the gap between data science, public health, and social impact — with a focus on creating inclusive, actionable solutions.

Dr Sam Hile

Department: Physics

Bio coming soon.

Dr Peter Wijeratne

Department: Informatics

Title: Probabilistic modelling of disease progression

Abstract: Disease progression models are special types of latent variable models that can infer trajectories of change in patients with a chronic degenerative condition, such as Alzheimer's disease. They provide unique insight into disease biology and staging systems with individual-level clinical utility. Here I will talk about some methods, applications, benefits and limitations of these approaches, and avenues for future research and collaboration.

Bio: Dr. Wijeratne has a background in physics, engineering, and computer science, which he uses to make machine learning models, primarily for healthcare. His current research interests include developing new probabilistic models of neurodevelopment and neurodegeneration. Along with Dr. Ivor Simpson, he leads the Learners of Interpretable Latent Information (LILI) lab in the Department of Informatics.

Dr Elizabeth Rendon-Morales

Department: Engineering

Dr. Elizabeth Rendon-Morales is an Associate Professor in Engineering in the Department of Engineering and Design at the University of Sussex. She sits on the EPSRC e-Futures electronics network steering group, the Sussex Research Development Concordant steering group and the EDI committee at the University of Sussex.

Dr. Rendon-Morales is an engineer and a researcher in the areas of design and development of sensors and robotics systems including the integration and experimental evaluation of medical technology and instrumentation. She has more than twenty years of experience from both academia and Tech industry. Her current research is focused on the design and development of novel flexible sensors and actuators systems to contribute to the next generation of on-demand drug delivery systems, as well as the design of CMOS sensors, machine vision systems and its integration in robotics for high precision applications in cardiac surgery including Transcatheter Aorta Valve Implantation (TAVI) and neurosurgery.

Early Afternoon Session

Aniruddha Girish Aramanekoppa

Probing the Stellar Populations of Distant Galaxies using the Balmer Break: Validating Galaxy Formation Models

In this investigation, the observed Balmer break strengths of distant galaxies are compared with predictions from galaxy formation models. The Balmer break is a spectral discontinuity that occurs at the Balmer limit (3645 Å) caused by the ionisation of the n=2 electrons. This feature is strongest in the spectra of A-type stars, making it an age sensitive feature, and thus a highly useful diagnostic for probing the stellar population of a galaxy. The break strength is strongly dependent on the age of the stellar population but can also be influenced by other physical properties, such as the metallicity or the star formation history. The recent deployment of the James Webb Space Telescope has made it possible to observe the Balmer break for galaxies in the early Universe. In this project, the Balmer break strengths of high-redshift galaxies were measured using the data collected by the NIRSpec instrument on the James Webb Space Telescope. These measurements were then compared with predictions from galaxy formation models. In this study, the predictions were obtained from the First Light and Reionization Epoch Simulations (FLARES). The results showed a reasonably good agreement with the FLARES predictions, as most of the measurements lie within the predicted range. The observed break strengths exhibit a broader distribution than the predictions. There is also a systematic shift towards smaller Balmer breaks in the observational data. The predicted medians were found to be between 3 and 5 standard deviations higher than the observed medians. This shift can be explained either by the high-redshift galaxies having lower than expected Balmer breaks, or through the existence of a selection bias in the dataset that favours younger stellar populations. Several possible extensions to this investigation are suggested, including extending the measurements to other datasets and comparing the results with other galaxy formation models.

Charlotte Robinson

Artificial Intelligence and Animal Behaviour

Abstract coming soon.

Nay Newman

Ant Visual Navigation

Many ant species are known to rely on visual information for navigation during foraging trips, particularly if other navigational tools are in conflict or fail. This has been shown experimentally in the field with desert ant species (Cataglyphis sp. & Myrmecia sp.) and wood ant species (Formica rufa). These animals are able to navigate through harsh desert or complex forest environments efficiently and effectively with very small brains to process sensory information and execute appropriate movements; as well as low resolution eyes. This project aims to explore the difficulty of a vision based navigation task with limited visual input, via image resolution, and limited computational capacity, via CNN model size, from an ant's eye perspective.

Renhui Ying

Applying computer vision tools in monitoring outdoor animal activities

Tracking animal activity in outdoor farm environments is crucial for livestock management, yet it remains a challenging task due to dynamic changes in cow appearance, variable lighting and unpredictable animal movements. Traditional vision-based systems, while effective indoors, often fail outdoors as they rely on consistent visual cues, leading to unstable tracking and poor identity association. This work introduces TwinTrack, a multi-object tracking framework designed to address these challenges. The proposed framework leverages a Twin-Level Contextual Feature Synthesizer (TLCFS) to extract both fine-grained visual details and high-level semantic features, ensuring robustness under diverse environmental conditions. Additionally, a Dynamic Long-Term Temporal Consistency Module (DLTC) improves tracking stability by mitigating the effects of dynamic behaviors and scene fluctuations. The application of TwinTrack to outdoor farm environments demonstrates its ability to monitor livestock effectively, with experimental results showing stable, long-term tracking performance.

Trevor Hewitt

The geometry of visual hallucinations

Visual hallucinations are commonly reported during psychedelic experiences, altered states of consciousness, psychiatric conditions, and neurological disorders, yet their contents remain poorly characterized quantitatively. The present study demonstrates that participants can reliably recreate visual hallucinations in a format suitable for quantitative computer-vision analysis. Simple visual hallucinations of geometric forms comparable to psychedelic and clinical hallucinations can be rapidly induced via stroboscopic light stimulation, representing an ideal paradigm for controlled quantitative analyses. To leverage this towards developing a quantitative understanding of the contents of visual hallucinations, two experiments were conducted in which over 100 participants were trained to recreate images of geometric visuals from hallucinatory and veridical visual experiences either through freehand drawing (Experiment 1) or a generative image-recreation interface (Experiment 2). Quantitative analysis revealed systematic differences in hallucinated geometric forms across strobe frequencies. These findings support the theorised effect of stroboscopic frequency on the content of the induced hallucinations while challenging current models of their neural mechanisms. Methods were validated with trials where participants recreated image stimuli instead of hallucinations. Image recreation is shown to be a viable tool for quantitative investigations into the phenomenology of visual hallucinations. A better understanding of visual hallucinations could shed light on how the brain constructs conscious experiences from sensory input. This research opens the door for computer-vision-based analyses of visual experiences across contexts, including psychedelic experiences and neurological disorders.

Maryam Seraj

Development of Haptic Interfaces for Medical Robots

Teleoperated medical procedures, such as tissue cutting, ultrasound scanning, palpation, etc., require precise motion control and robust interaction with complex, uncertain environments. Delays in communication, unpredictable environments, limited operator experience, and a mismatch between human hand precision and the capabilities of robotic execution are common problems for these processes. This mismatch is particularly evident in commercial teleoperated robots, which often exhibit high inertia, limited actuation power, and low natural frequencies; factors that reduce their responsiveness and accuracy during delicate operations. As a result, task precision, repeatability, and overall system reliability are all decreased. Furthermore, employing compliant controllers such as impedance control presents new issues. While impedance control increases safety by allowing the robot to adapt to contact forces, it also leads to path deviations in directions where the robot's manipulability is poor. For example, during tissue cutting, limited actuation capability in certain directions may lead the robot to deviate from a straight path, resulting in curved or inaccurate incisions. These issues get worse in teleoperation setups, where the operator lacks direct tactile feedback and may not intuitively compensate for robot limitations. To address these challenges, this study proposes a novel framework that integrates a Fractal Impedance Controller (FIC) strategy, an admittance controller, and a custom-designed, high-frequency haptic interface based on a Coaxial Spherical Parallel Mechanism (CSPM). By unifying FIC with this active, task-adaptive interface, the system achieves compliant interaction without compromising accuracy. Regardless of operator skill level, the end result is a more capable and intuitive teleoperated system that can consistently carry out critical medical procedures. This framework lays the groundwork for wider use in any situation requiring high-precision, dynamic human-robot collaboration, not just medical applications.

Late Afternoon Session

Jimena Berni

Analysis of neuronal activity with graph method

We investigated developmental changes in neuromotor activity patterns in Drosophila melanogaster larvae by combining calcium imaging with a novel graph-based mathematical framework. This allows to perform relevant quantitative comparisons between first (L1) and early third (L3) instar larvae. We found that L1 larvae exhibit higher frequencies of spontaneous neural activity that fail to propagate, indicating a less mature neuromotor system. In contrast, L3 larvae show efficient initiation and propagation of neural activity along the entire ventral nerve cord (VNC), resulting in longer activity chains. The time of chain propagation along the entire VNC is shorter in L1 than in L3, probably reflecting the increased length of the VNC. On the other hand, the time of peristaltic waves through the whole body during locomotion is much faster in L3 than in L1, so correlating with higher velocities and greater dispersal rates. Hence, the VNC-body interaction determines the characteristics of peristaltic waves propagation in crawling larvae. Further, asymmetrical neuronal activity, predominantly in anterior segments of L3 larvae, was associated with turning behaviors and enhanced navigation. These findings illustrate that the proposed quantitative model provides a systematic method to analyze neuromotor patterns across developmental stages, for instance, helping to uncover the maturation stages of neural circuits and their role in locomotion.

Giuseppe Castiglione

Deep Learning Theory for Learning Dynamics

Deep neural networks trained with gradient descent exhibit varying rates of learning for different patterns. However, the complexity of fitting models to data makes direct elucidation of the dynamics of learned patterns challenging. To circumvent this, many works have opted to characterize phases of learning through summary statistics known as order parameters. In this work, we propose a unifying framework for constructing order parameters based on the Neural Tangent Kernel (NTK), in which the relationship with the data set is more transparent. In particular, we derive a local approximation of the NTK for a class of deep regression models (SIRENs) trained to reconstruct natural images. In so doing, we analytically connect three seemingly distinct phase transitions: the emergence of wave patterns in residuals (a novel observation), loss rate collapse, and NTK alignment. Our results provide a dynamical perspective on the observed biases of SIRENs, and deep image regression models more generally.

Enrico Caprioglio

Synergistic information and network science

High-order interdependencies are central features of complex systems, yet a mechanistic explanation for their emergence remains elusive. Currently, it is unknown under what conditions high-order interdependencies, quantified by the information-theoretic construct of synergy, emerge in systems governed by pairwise interactions. For linear Gaussian systems of arbitrary dimension, we prove that antibalanced correlational structures are sufficient for synergy-dominance and that antibalanced interaction motifs in Ornstein-Uhlenbeck processes are necessary for synergy-dominance. We validate the applicability of these analytical insights in Ising, oscillatory, and empirical networks. Our results demonstrate that pairwise interactions alone can give rise to synergistic information in the absence of explicit high-order mechanisms, and highlight structural balance theory as an instrumental conceptual framework to study high-order interdependencies.

Yidan Xu

Continuous Assessment of Fear of Falling in Parkinson's Disease Using Wearable Sensors and Machine Learning

Fear of falling (FoF) is a prevalent and debilitating emotional experience in Parkinson's disease (PD), affecting 37%-59% of people with PD. While a certain level of FoF can promote safety, maladaptive patterns may emerge when self-perceived FoF does not align with actual balance ability. This mismatch can lead to either excessive avoidance, reduced mobility, and psychological distress, or to inappropriate risk-taking in individuals who underestimate their fall risk. Despite its importance, FoF is typically assessed through retrospective questionnaires at infrequent clinic visits, which fail to reflect fluctuations in daily life.

This research investigates a continuous, real-world approach to assessing FoF using wearable sensors and machine learning. A new dataset will be collected from people with PD in their homes, capturing multi-day body movement data (e.g., from accelerometers and gyroscopes) alongside various self-reported emotional labels including FoF level over time.

The study focuses on extracting a range of features, including statistical and signal-based metrics, gait variability, turning characteristics, freezing-related indicators, and movement complexity measures, to identify movement patterns that are relevant for the automatic assessment of fear of falling in people with Parkinson's disease. To address the limitation of small and scarce datasets, transfer learning methods will be used by leveraging related datasets from other people with PD and healthy populations. The generalisability of models across subgroups (e.g., disease stage, gender) and across settings (lab vs home) will also be explored.

This work aims to support personalised, objective monitoring of fear of falling in PD, enabling early intervention and better self-management in everyday life.

'Interdisciplinary: Improving Dialogue and Collaboration' Panel

Prof Thomas Nowotny

Head of Sussex AI

Bio coming soon.

Dr Aline Amorim Graf

Research Fellow in Materials Physics

Dr Aline Amorim Graf is a Research Fellow in the Materials Physics group. Her current work involves developing and processing nanomaterial inks designed to tune and exploit their enhanced electrical and thermal properties. As an Early Career Researcher, she also worked briefly in the Engineering department, modelling thermal properties of an office and the relationship with its occupant's comfort. She received her PhD from the University of Sussex for her work on extracting reliable and statistically significant information from Raman spectroscopy of two-dimensional nanomaterials.

Dr Elizabeth McKenzie

Assistant Director of Strategy and Operations for the Sussex Centre for Quantum Technologies

Elizabeth leads on the Centre's Strategy and Operations, driving and implementing the activities that increase our research profile, engagement and partnerships and develop the researcher community.

Amber Shepherd

PhD Student, Ion Trap Cavity-QED and Molecular Physics (ITCM) group

Amber Shepherd is a PhD student at the University of Sussex, working in the Ion Trap Cavity-QED and Molecular Physics (ITCM) group. She is working on an experiment developing a quantum clock to test fundamental physics. She has experience of organising cross-departmental outreach initiatives including Soapbox Science Brighton and Portslade Science Fair, and co-founded the MPIE conference last year to bring together researchers from MPS, EngInf and the wider faculty.