Computing Science and Mathematics Seminars,
Autummn 2023
Unless otherwise state, seminars will take place in Room 4B96, Cottrell Building, University of Stirling from 13.00 to 14.00 on Friday afternoons, followed by informal discussions.
Autumn 2023
| Date | Speaker | Title/Abstract |
|---|---|---|
|
Friday 15 September |
No seminar |
|
| Friday 22 September |
No seminar |
|
| Friday 29 September |
No seminar |
|
| Friday 6 October |
Dylan Powell, Health Sciences, University of Stirling | Measuring brain health one step at a time. How wearables and digital technology are transforming the diagnosis and monitoring of neurological conditions Welcome to this guest lecture on ‘Measuring brain health one step at a time’ where we delve into the world of wearable technology and its emerging role in healthcare. In this talk Dr Dylan Powell explores how wearables are redefining the way we monitor and manage health conditions. In this lecture, Dylan will introduce the technical considerations for using wearables and their increasing prevalence in our daily lives. How these devices, ranging from research grade devices & smartwatches, have evolved from ‘gadgets’ to power tools for monitoring various aspects of health. Dr Powell will explore how wearables are improving the diagnosis and management of mild Traumatic Brain Injury (mTBI), Parkinsons and Dementia by tracking vital signs and analysing subtle movements and gait patterns. Throughout the lecture, Dr Powell will highlight the wealth of data generated by wearables and the opportunities and challenges associated with managing and interpreting this data. We'll conclude by emphasising the broader impact of wearable technology in healthcare, where early detection and proactive health management are becoming increasingly achievable.
Speaker bio: Dylan Powell is a Lecturer in Public Health & Innovation at the University of Stirling and recently appointed Editor (News & Views) at Nature Digital Medicine. Dylan’s professional background as a clinician (Physiotherapist) and researcher has spanned the NHS, professional sport, and within industry (Deloitte). His PhD in Computer Science at Northumbria University (under Dr Alan Godfrey) explored the use of wearables and digital biomarkers across a variety of health conditions, including sports-related concussion. Dylan most recently was the Knowledge Sharing Lead for Deloitte and has experience delivering projects ranging from sustainability to digital transformation across Life Sciences & Healthcare. |
| Friday 13 October |
Varhid Akbari, Computing Science and Maths, University of Stirling | Advanced statistical modelling of polarimetric SAR data for land cover change detection analysis In this talk, I will present a determinant ratio test (DRT) statistic to measure the similarity of two covariance matrices for unsupervised change detection in polarimetric radar images. The multilook complex covariance matrix is assumed to follow a scaled complex Wishart distribution. In doing so, the distribution of the DRT statistic is analytically derived which is exactly Wilks’s lambda of the second kind distribution, with density expressed in terms of Meijer G-functions. Due to this distribution, the constant false alarm rate (CFAR) algorithm is derived in order to achieve the required performance. More specifically, a threshold is provided by the CFAR to apply to the DRT statistic producing a binary change map. Finally, simulated and real multilook polarimetric radar data are employed to assess the performance of the method and is compared with the Hotelling–Lawley trace (HLT) statistic and the likelihood ratio test (LRT) statistic.
Speaker bio: Dr Vahid Akbari is a Lecturer in Data Science/Artificial Intelligence at the University of Stirling. He holds a PhD degree in physics with a specialisation in radar remote sensing and data analytics from UiT-The Arctic University of Norway in Tromsø, Norway. His primary research interests revolve around the intersection of radar remote sensing and statistical modelling/data science, with a particular focus on the applications in environmental monitoring. |
| Friday 20 October |
Robert Mooney, Leonardo UK | An Introduction to Modern Electronic Warfare Airborne Electronic Warfare systems search for emission sources across the electromagnetic spectrum. Discovered sources are separated, characterised and identified in order to provide the aircraft with general situation awareness and give prompt warning of specific threats and opportunities. Modern systems must operate in an increasingly congested and contested environment, against increasingly agile and sophisticated software-defined threats, but are also required for far more than threat warning, with passive sensing offering many advantages over active radars in terms of range, stealth, bandwidth, and spatial coverage. Together, these issues pose significant challenges for both signal processing and control systems, but also significant opportunities, especially as next generation systems are being developed. This talk intends to give a broad overview of modern Electronic Warfare and our problem space, as an invitation to further discussion. Speaker bio: Dr Robert Mooney is a Research Engineer at Leonardo UK, working on signal processing and autonomy for Electronic Warfare applications. His academic background and PhD are in Theoretical High Energy Physics, specifically the structure of scattering amplitudes and related observables in Supersymmetric Gauge Theories. |
| Friday 27 October |
No seminar |
|
| Friday 3 November |
Sandy Carmichael, Computing Science and Maths, University of Stirling | Computer Vision and Signal Processing Methods for Analysing Atlantic Salmon Gill Histology Images One of the primary indicators of decline in Atlantic salmon gill health is hyperplasia, a condition involving abnormal cell growth. We have developed an innovative signal/image processing pipeline for analysing gill histology images, which evaluates image texture using a new signal processing technique, the empirical wavelet transform, in tandem with deep learning. Our approach identifies hyperplastic regions in whole-slide images, providing a fine-grained understanding of the lesion's prevalence. Furthermore, we have incorporated unsupervised learning with a variational autoencoder to identify regions of interest. As we advance this approach, it holds the potential to provide a quantitative computer-assisted hyperplasia score to support a histopathological diagnosis by humans. The outlined procedure can be adapted to evaluate other gill conditions and histopathological images beyond gills. Speaker bio: Sandy Carmichael is a PhD student at the University of Stirling and Scotland's Rural College (SRUC). He started his doctoral research in October 2019, shortly after he graduated with a BSc. (Hons) in Software Engineering from the University of Stirling. His research interests include computer vision, medical imaging, signal processing, and natural language processing, with a particular focus on their applications to aquatic animal health data. |
| Friday 10 November |
Jie Zhang, Department of Informatics, King's College London | LLM is Like a Box of Chocolates: the Non-determinism of ChatGPT in Code Generation There has been a recent explosion of research on Large Language Models (LLMs) for software engineering tasks, in particular code generation. However, results from LLMs can be highly unstable; nondeterministically returning very different codes for the same prompt. Non-determinism is a potential menace to scientific conclusion validity. When non-determinism is high, scientific conclusions simply cannot be relied upon unless researchers change their behaviour to control for it in their empirical analyses. In this talk, I will introduce our empirical study which demonstrates that non-determinism is, indeed, high, thereby underlining the need for this behavioural change. We choose to study ChatGPT because it is already highly prevalent in the code generation research literature. We report results from a study of 829 code generation problems from three code generation benchmarks (i.e., CodeContests, APPS, and HumanEval). Our results reveal high degrees of non-determinism: the ratio of problems with zero equal test output among code candidates is 72.73%, 60.40%, and 65.85% for CodeContests, APPS, and HumanEval, respectively. In addition, we find that setting the temperature to 0 does not guarantee determinism in code generation, although it indeed brings less non-determinism than the default configuration (temperature=1). These results confirm that there is, currently, a significant threat to scientific conclusion validity. In order to put LLM-based research on firmer scientific foundations, researchers need to take into account non-determinism in drawing their conclusions. In addition to this work on code generation, I will also briefly introduce my recent work “Large Language Models in Fault Localisation” and our survey paper “Large Language Models for Software Engineering: Survey and Open Problems”. Speaker bio: Dr Jie M. Zhang is a lecturer of computer science at the Department of Informatics in King’s College London. Before joining King’s she was a Research Fellow at University College London and a research consultant for Meta. She got her PhD degree from Peking University in 2018. Her main research interests are software testing, machine learning testing, machine learning trustworthiness, and code generation. She has published over 20 papers in top-tier venues including ICLR, ICSE, FSE, ASE, ISSTA, TSE, and TOSEM since her PhD started in 2012. She is a steering committee member of ICST and the steering committee chair of SBST. She is a co-chair of Internetware 2024, ICST 2024 Journal-first track, ASE 2023 NIER track, SANER 2023 Journal-First Track, PRDC 2023 Fast Abstract Track, SBST 2021, Mutation 2021&2020, and ASE 2019 Student Research Competition. Over the last three years, she has been invited to give over 20 talks at conferences, universities, and IT companies, including two keynote talks. She has also been invited as a panellist for several seminars on large language models. She has been selected as one of the fifteen 2023 Global Chinese Female Young Scholars in interdisciplinary AI by Baidu. |
| Friday 17 November |
Rebecca Moussa, Computer Science, University College London | Multi-objective Ensemble Generation (MEG) Recent studies have found that ensemble prediction models (i.e., aggregation of multiple base classifiers) can achieve more accurate results than those that would have been obtained by relying on a single classifier when used to make predictions. However, designing an ensemble requires a non-trivial amount of effort and expertise with respect to the choice of the set of base classifiers, their hyper-parameter tuning, and the choice of the strategy used to aggregate the predictions. An inappropriate choice of any of these aspects can lead to over- or under-fitting, thereby heavily worsening the performance of the prediction model. Examining all possible combinations is not computationally affordable, as the search space is too large, and there is a strong interaction among these aspects, which cannot be optimised separately. In this talk, I will present our novel approach which leverages the power of multi-objective evolutionary search to automatically produce prediction ensembles. We refer to our proposal as (M)ulti-objective (E}nsemble (G)eneration (MEG). We verify the effectiveness of MEG in detecting defects across software versions as well as in detecting a new type of defect: Gender, race and age bias in software systems. This talk will cover MEG as well as its application in Software Defect Prediction and Software Fairness. Speaker bio: Rebecca Moussa is a Post Doctoral Researcher in the Department of Computer Science at University College London (UK), where she is a member of the SSE, SOLAR and CREST groups. She received her PhD from the University College London, under the supervision of Prof Federica Sarro and Prof Mark Harman. Her main expertise lies in automated approaches to software project management and software quality leveraging the power of software analytics and search-based software engineering. |
| Friday 24 November |
Leonardo Bezerra, Computing Science and Maths, University of Stirling | Promoting and sustaining accountability in artificial intelligence applications
Technology has been the catalyst for major revolutions societies have gone through, and each new revolution brings social challenges that governments must address. In turn, regulation acts as a form of feedback that directs how the breakthrough technology of the time will have to be adapted. Currently, the most pressing technology revolution is being powered by social media, big data, and artificial intelligence (AI). Though this revolution has been taking place for over a decade now, recent years have seen an astounding increase in the pace with which these applications are being developed and deployed. Not surprisingly, regulatory agencies around the world have been unable to cope with this speed and have just recently started to move from a data-centred to an AI-centred concern. More importantly, governments are still beginning to mature their understanding of AI applications in general, let alone discuss AI ethics and how to promote and sustain accountability in AI applications. In turn, companies that use AI in their applications have also begun to display some public level of awareness, even if often vague and not substantiated by concrete actions. In this talk, we will briefly overview efforts and challenges regarding AI accountability and how major AI players are addressing it. The goal of the talk is to stir future project collaborations from a multidisciplinary perspective. Speaker bio: Dr Leonardo Bezerra joined the University of Stirling as a Lecturer in Artificial Intelligence (AI)/Data Science in 2023, after having been a Lecturer in Brazil for the past 7 years. He received his Ph.D. degree from Université Libre de Bruxelles (Belgium) in 2016, having defended a thesis on the automated design of multi-objective evolutionary algorithms. His research experience spans from applied data science projects with public and private institutions to supervising theses on automated and deep machine learning. Recently, his research has concentrated on the social impact of AI applications, such as disinformation through social media recommendation algorithms and the disruptive potential of generative AI. |
| Friday 1 December |
Piotr Teodorowski, FHSS Management and Support, University of Stirling | Public involvement in big data research Public involvement became more popular at every stage of research. NIHR defines it as “research being carried out ‘with’ or ‘by’ members of the public rather than ‘to’, ‘about’ or ‘for’ them.” Involving members of the public in research assists researchers in ensuring that public views are taken into consideration and that research is person-centred and relevant to the public. Funders might require it at the application stage, and some journals would not publish papers without public involvement in research. However, so far public involvement has been more popular in qualitative rather than quantitative research. One way to explain this is that public involvement in quantitative research can face additional challenges. This might be particularly the case in big data research due to its complexity. This seminar will focus on how to involve members of the public in big data research. This will be discussed from the perspective of researchers and the public. Speaker bio: Piotr Teodorowski is a health services researcher at the Faculty of Health Science and Sport, University of Stirling. His research interests include public involvement and engagement. Piotr’s doctoral research explored how to involve and engage the public (especially seldom-heard communities) in big data research. He is a member of the working group in the Public Engagement in Data Research initiative (PEDRI) that creates new Public Involvement and Engagement Best Practice Standards for the Use of Data for Research and Statistics. |
Computing Science and Mathematics
Faculty of Natural Sciences
University of Stirling
Stirling
FK9 4LA
Scotland
UK