School of Computing Science

Grounded radiology report generation with large multimodal models

Group: Computing Technologies for Healthcare
Speaker: Stephanie Hyland, Microsoft
Date: 22 November, 2024
Time: 14:00 - 15:00
Location: 423 SAWB

 Abstract: 

Machine learning approaches for automated radiology report generation show increasing promise for assisting radiologists in their work. However, for such systems to have clinical impact they must exhibit both strong performance and usability. We demonstrate that performance can be improved by providing models with more realistic reporting context akin to that used by the radiologist, such as the prior study, indication for the study, and lateral view (for chest X-ray reporting). To improve usability, we task the model with generating not only a narrative description of findings, but also localising each finding on the image – a task we call grounded radiology report generation. In this talk, I will describe this novel task, the model we designed to tackle it ("MAIRA-2"), and a new LLM-based evaluation framework ("RadFact") we designed for radiology report generation both with and without grounding.
MAIRA-2 is available here and RadFact here.  

Biography: As of 2019, Stephanie is a senior researcher at Microsoft Research Cambridge in the Health Intelligence group. She works on machine learning for healthcare. In 2018, she received a PhD in Computational Biology and Medicine from the Tri-Institutional PhD Program of Cornell University. During her PhD she was situated at the Sloan Kettering Cancer Center in NYC, and later ETH Zürich. Prior to that, she studied theoretical physics at Trinity College Dublin, and applied mathematics (Part III) at the University of Cambridge.

This event is organised by the Biomedical AI and Imaging Lab.

Grounded radiology report generation with large multimodal models

Group: Computing Technologies for Healthcare
Speaker: Stephanie Hyland, Microsoft
Date: 22 November, 2024
Time: 14:00 - 15:00
Location: 423 SAWB

 Abstract: 

Machine learning approaches for automated radiology report generation show increasing promise for assisting radiologists in their work. However, for such systems to have clinical impact they must exhibit both strong performance and usability. We demonstrate that performance can be improved by providing models with more realistic reporting context akin to that used by the radiologist, such as the prior study, indication for the study, and lateral view (for chest X-ray reporting). To improve usability, we task the model with generating not only a narrative description of findings, but also localising each finding on the image – a task we call grounded radiology report generation. In this talk, I will describe this novel task, the model we designed to tackle it ("MAIRA-2"), and a new LLM-based evaluation framework ("RadFact") we designed for radiology report generation both with and without grounding.
MAIRA-2 is available here and RadFact here.  

Biography: As of 2019, Stephanie is a senior researcher at Microsoft Research Cambridge in the Health Intelligence group. She works on machine learning for healthcare. In 2018, she received a PhD in Computational Biology and Medicine from the Tri-Institutional PhD Program of Cornell University. During her PhD she was situated at the Sloan Kettering Cancer Center in NYC, and later ETH Zürich. Prior to that, she studied theoretical physics at Trinity College Dublin, and applied mathematics (Part III) at the University of Cambridge.

This event is organised by the Biomedical AI and Imaging Lab.

The 35th British Machine Vision Conference

Group: Scottish Informatics and Computer Science Alliance (SICSA)
Speaker: SICSA Event, SICSA
Date: 25 November, 2024
Time: 00:00 - 00:00
Location: Scottish Event Campus, Scottish Event Campus, Glasgow, United Kingdom

The British Machine Vision Conference (BMVC) is the British Machine Vision Association’s (BMVA) annual conference on machine vision, image processing, and pattern recognition. It is one of the major international conferences on computer vision and related areas held in the UK. With increasing popularity and quality, it has established itself as a prestigious event on the vision calendar. Find out more details on the MBVC 2024 Conference Website.

When LLMs Meet Recommendations: Scalable Hybrid Approaches to Enhance User Experiences

Group: Information Retrieval (IR)
Speaker: Jianling Wang, Google DeepMind
Date: 25 November, 2024
Time: 15:00 - 16:00
Location: Sir Alwyn Williams Building, 422 Seminar Room

Title:
When LLMs Meet Recommendations: Scalable Hybrid Approaches to Enhance User Experiences

Bio:
Jianling Wang is a senior research scientist working at Google DeepMind. She obtained her Ph.D. degree from the Department of Computer Science and Engineering at Texas A&M University, advised by Prof. James Caverlee. Her research interests generally include data mining and machine learning, with a particular focus on recommendation systems and graph neural networks.

Abstract:
While LLMs offer powerful reasoning and generalization capabilities for user understanding and long-term planning in recommendation systems, their latency and cost hinder direct application in large-scale industrial settings. The talk will cover our recent work on scalable hybrid approaches that combine LLMs and traditional recommendation models. We'll explore their effectiveness in tackling challenges like cold-start recommendations and enhancing user exploration.

Introduction to the Raspberry Pi Computing Education Research Centre

Group: Education and Practice Research Section
Speaker: Dr Sue Sentance, University of Cambridge
Date: 25 November, 2024
Time: 16:00 - 17:00
Location: Boyd Orr Building Room 711

At the University of Cambridge, we established a research centre focusing on computing and AI education for young people in late 2021. A joint initiative with the Raspberry Pi Foundation, the Centre shares its mission to support young people in realising their full potential through the power of computing and digital technologies. The Centre is developing projects linked to six themes: Broadening Participation, AI Education, Programming Education, Physical Computing, Curriculum and Pedagogy, including Teacher Professional Development. Our approach is to work on topics that are directly relevant to practice, and to work with educators and other practitioners through participatory design methods.  In this talk I'll describe the main projects we've been working on, and I hope we'll be able to find areas where there is scope for collaboration with the work being conducted in Glasgow.

Biography

Sue Sentance is Director of the Raspberry Pi Computing Education Research Centre at the University of Cambridge, UK. She received her PhD in Artificial Intelligence from Edinburgh University in 1993, and since then has worked as a secondary teacher, teacher trainer, university lecturer and researcher. She was the Chief Learning Officer for the Raspberry Pi Foundation from 2018-2023, and in that period played a leading role in the DfE-funded National Centre for Computing Education. Sue's research interests include programming pedagogy, particularly PRIMM, teacher professional development, AI education and culturally relevant pedagogy, and has edited a textbook for prospective teachers and researchers: Computer Science Education: Perspectives on Teaching and Learning in School.

Systems Teaching Discussion

Group: Systems Seminars
Speaker: Dr Jeremy Singer
Date: 26 November, 2024
Time: 14:00 - 15:00
Location: SAWB 422 and online -- https://uofglasgow.zoom.us/j/86057258886?pwd=y2kOP3VMnErustCF9dtN6hHIVZ4VvH.1

Systems topics form a core spine of our undergraduate Computing Science curriculum. In this open discussion forum, we will explore the current systems topics that are mandatory for all students and the options available for honours students. We will examine the connections between our courses, identifying areas for rationalization. We will also consider possible new topics to be introduced, particularly in light of our new Cybersecurity masters programme.

SICSA AI Online Workshop with Leo Porter

Group: Scottish Informatics and Computer Science Alliance (SICSA)
Speaker: SICSA Event, SICSA
Date: 26 November, 2024
Time: 16:00 - 17:30
Location: TBA

On Tuesday 26 November 16:00-17:30 SICSA welcomes Professor Leo Porter, co-author of “Learn AI-Assisted Python Programming” and Associate Teaching Professor at UC San Diego. Leo will present ideas from his book which ‘teaches you everything you need to start programming Python in an AI-first world’. (Published by Manning, September 2023). There will also be a Q&A session. In addition, we will host some in-person meetings to view the talk and participate in follow-up discussions. Register for the workshop or in-person meeting with follow-up discussion The locations for the in-person meetings are as follows: University of Glasgow Room 422, Sir Alwyn Williams Building University of Edinburgh Room 8.02, Appleton Tower Edinburgh Napier University Room E11, Merchiston Campus If you would like to organise any additional meetings locally or have any other questions, please contact us – sicsa@glasgow.ac.uk SICSA Talks Series This workshop kicks off a series of SICSA talks/discussions where we will consider how machine learning (AI and LLMs) can beneficially impact our teaching and learning practice. The aim of the discussion following Leo’s talk is to set the agenda for the series. The series will draw on experience at SICSA institutions and invited speakers. The series will include: • Learning programming with co-pilots. • System development using Machine Learning components. • System development using Machine Learning assistants. • Software Engineering in the context of Machine Learning. • Supporting students to make effective use of Machine Learning tools in coursework. Please let us know if you have any ideas for additional workshops by contacting us – sicsa@glasgow.ac.uk

Upwards x GWiCS seminar: Effective Time, People, and Project Management

Group: School of Computing Science
Speaker: Professor Simone Stumpf and Dr Matthew Barr, University of Glasgow
Date: 27 November, 2024
Time: 13:00 - 14:00
Location: SAWB 423, Sir Alwyn Williams Building

Topic: Effective Time, People, and Project Management

Speakers:

- Professor Simone Stumpf (Computing, GIST Section lead)

- Dr Matthew Barr (Computing, EAP Section lead)

Location:

- in Room SAWB 423

- and on Zoom (to join remotely): https://uofglasgow.zoom.us/j/83133204070?pwd=pGhpQIgXyjebI0xKhRv0U4tI5ZF6Qb.1

This is a joint Upwards and GWiCS seminar.

 
What is Upwards?
 
Upwards is the School’s research culture seminar, covering all facets of developing, conducting, and disseminating research and related topics (e.g. managing a research team, time management to do research, connections between research and teaching). It is open to everyone in the School, but a specific aim is to support ECR development and some sessions are aimed mainly at PGRs and/or PDRAs.

How are the seminars held?

Upwards seminars are held in person in the School to bring people together. However, the sessions are also streamed on Zoom to allow people to join remotely, if they cannot attend in person. The seminars are not recorded and slides are not shared to preserve the off-the-record atmosphere of the seminars, which allows speakers to share personal experiences.

What will this session be about?
 
It is up to the speakers to set the agenda for their Upwards talks, but the idea for this joint Upwards/GWiCS seminar is to hear, for example, about best practices and useful strategies for: how to organise one's weekly and daily schedule, how to protect time set aside for research development, how to manage postdocs and PhDs as their supervisor, how to keep small and large research projects on track, and how to keep a healthy balance between the many tasks one has to do and life outside of the job.

Peer Interaction (PI), and the interplay of different types of research on it

Group: Centre for Computing Science Education (CCSE)
Speaker: Steve Draper, University of Glasgow
Date: 28 November, 2024
Time: 15:00 - 16:00
Location: F121 Lilybank Gardens and Online

"Peer Interaction" (PI) refers to the important ways that a learner interacting with not a teacher but a fellow student or "peer" can be more productive of learning than interacting with a teacher one-to-one.

There is a long history of publications on peer interaction (PI) in education (although the earliest I will cite is 1982, important points were published in the 19th century and possibly before).  Some of this literature was written, and still more has been used and explored, by CCSE members.  PI is important and applicable in all disciplines.  It has been used quite extensively in computer science, and some of the most powerful uses have been in Maths and in Physics.

One purpose of this talk is to frame education research as to do with universal theories that apply across disciplines, rather than seeing it as a craft that works differently in different disciplines.  A second aim is not only to introduce and discuss some of this literature, but also to show the different kinds of research PI has attracted, each with different merits.  (Another seminar might care to discuss how we could or should combine some of these types of research into a better overall program of research on one learning design or one course.)

The talk outlines five contrasting types of contribution to PI in the literature:

1. Insights from theory, suggesting how and why PI can produce better quality and quantity of learning.
2. The variety of successful Learning Designs using PI (examples of different course or lesson plans)
3. The best quantitative evidence of learning gains from PI
4. The wider importance of PI in education and especially in HE:  pervasive, but seldom studied or measured.
5. Radical PI:  Connectivism, and approaches where learners, not teachers, decide the curriculum and the learning activities.

Collaborative Opportunities with Singapore Institute of Technology

Group: Computing Technologies for Healthcare
Speaker: Ng Pai Chet, https://www.singaporetech.edu.sg/directory/faculty/pai-chet-ng
Date: 29 November, 2024
Time: 11:00 - 12:00
Location: 422 SAWB - https://uofglasgow.zoom.us/j/84369571835?pwd=bWTbM95W8lWzGK6NsJZltvUNLoOJT2.1

This event will focus on the joint initiatives between the School of Computing Science (SoCS) at the University of Glasgow and the Singapore Institute of Technology, particularly the newly established Glasgow-Singapore Scholarships.

The School of Computing Science is offering two Glasgow-Singapore Joint PhD Scholarships. These prestigious awards will support doctoral researchers in undertaking collaborative projects, with each scholar spending 18-24 months in Singapore. This unique opportunity combines the expertise of both institutions while providing international research experience.

Biography: 

I am currently an Assistant Professor at the Singapore Institute of Technology (SIT). Prior to this role, I served as a Postdoctoral Fellow at the Department of Electrical and Computer Engineering, University of Toronto. I obtained my Ph.D. degree from the Hong Kong University of Science and Technology.

During my time as a Postdoctoral Fellow, I actively collaborated with Ph.D. students and industry research scientists, spearheading diverse projects ranging from human activity recognition to human stress detection. As a project manager, I undertook the responsibilities of coordinating meetings, assimilating research inputs, and providing progress updates to sponsors.

My research during my Ph.D. studies yielded over 10 publications in top-tier IEEE journals and conferences, including the IEEE IoT Journal, the IEEE Transactions on Mobile Computing, and the IEEE Transactions on Wireless Communications. My research pursuits are centered around the application of artificial intelligence techniques to hyperspectral imaging on consumer devices, applied AI for applications in Internet of Things (IoT) and physiological signal processing, multimodal sensing and federated learning, wireless positioning and sensing, and human-robot interaction.

Organisation: This event is organised by the Biomedical AI and Imaging Group and the Research Student Committee

ACI 2024 - The Eleventh International Conference on Animal-Computer Interaction

Group: Scottish Informatics and Computer Science Alliance (SICSA)
Speaker: SICSA Event, SICSA
Date: 02 December, 2024
Time: 00:00 - 00:00
Location: TBA

ACI is the leading International Conference on Animal-Computer Interaction. It is a highly multidisciplinary event drawing researchers and practitioners from diverse backgrounds to share and discuss work and topics related to the research and design of computing-enabled and interactive technology for and with animals. We are keen to be as inclusive as possible. We wish to welcome a wide range of contributions and participants to the conference, promote a constructive dialogue around the animal-centred research and design of computing-enabled systems, and foster the development of ACI as a discipline. The conference is open to contributions from researchers and practitioners in a wide range of fields, including (but not limited to) ethics, behavior analysis, psychology, veterinary behavior, zoology, ethology, interaction design, computer science, and electrical engineering. Find out more about the conference and submission deadlines.

1st International Workshop on Low carbon Computing (LOCO 2024)

Group: Scottish Informatics and Computer Science Alliance (SICSA)
Speaker: SICSA Event, SICSA
Date: 03 December, 2024
Time: 00:00 - 00:00
Location: Advanced Research Centre (ARC), 11 Chapel lane, Glasgow, G11 6EW, United Kingdom

The 1st International Workshop on Low Carbon Computing (LOCO 2024) will bring together researchers and practitioners with a keen interest in low carbon and sustainable computing. The workshop will provide a forum for sharing new ideas, for presenting ongoing work and early results, as well as for bringing forward well-founded criticism. LOCO 2024 is an initiative of the Scottish Programming Languages Institute (SPLI), supported by the Scottish Informatics and Computer Science Alliance (SICSA), and was inspired by the Programming for the Planet (PROPL) workshop. View more information and register online.

1st International Workshop on Low Carbon Computing (LOCO 2024)

Group: Scottish Informatics and Computer Science Alliance (SICSA)
Speaker: SICSA Event, SICSA
Date: 03 December, 2024
Time: 00:00 - 00:00
Location: Advanced Research Centre (ARC), 11 Chapel lane, Glasgow, G11 6EW, United Kingdom

The 1st International Workshop on Low Carbon Computing (LOCO 2024) will bring together researchers and practitioners with a keen interest in low carbon and sustainable computing. The workshop will provide a forum for sharing new ideas, for presenting ongoing work and early results, as well as for bringing forward well-founded criticism. LOCO 2024 is an initiative of the Scottish Programming Languages Institute (SPLI), supported by the Scottish Informatics and Computer Science Alliance (SICSA), and was inspired by the Programming for the Planet (PROPL) workshop. View more information and register online.

1st International Workshop on Low Carbon Computing (LOCO)

Group: Low Carbon and Sustainable Computing
Speaker: Anne Currie, Ayse Cuskun, and others
Date: 03 December, 2024
Time: 09:00 - 18:00
Location: 423 SAWB

GLACSIL Industrial Studentship Showcase

Group: Glasgow Computing Science Innovation Lab
Speaker: TBC
Date: 03 December, 2024
Time: 14:00 - 17:00
Location: Advanced Research Centre 237C

At this event we'll celebrate the work and achievements of the industrial research students sponsored by GLACSIL partners via talks, technology demonstrations, plus festive drinks and nibbles. 

Further details will be added in due course. 

[FATA Seminar] Runtime Instrumentation for Reactive Components

Group: Formal Analysis, Theory and Algorithms (FATA)
Speaker: Duncan Paul Attard, University of Glasgow
Date: 03 December, 2024
Time: 15:00 - 16:00
Location: Sir Alwyn Williams Building, 422 Seminar Room

Software instrumentation is central to many forms of program analyses. Runtime verification is one such prominent analysis example that imposes stringent constraints on the software behaviour reported by an instrumentation. Reactive components, ubiquitously found in many software systems, dictate further requirements on the instrumentation that is applied to them, such as the need for responsiveness and elasticity. In addition, the heterogeneity of their constituent components often relegates the runtime analysis to outline monitors, which execute separately from system components.

This talk discusses the challenges faced when reactive component systems are instrumented using decentralised outline monitors. We present a novel decentralised instrumentation algorithm based on next-hop routing for this setting, arguing that it satisfies the requirements for monitoring reactive component systems. We also discuss its performance based on an empirical evaluation and show that the runtime behaviour of our algorithm makes it feasible for soft real-time reactive applications. Compared to inline and centralised instrumentation, we conclude that our decentralised approach induces comparable latency to inline monitoring for software that performs long-running, computationally-intensive tasks, such as in Big Data stream processing.

This is joint work with Adrian Francalanza, Luca Aceto and Anna Ingolfsdottir from the University of Malta, Gran Sasso Science Institute and Reykjavik University, respectively.

-----------------------------------

This event is part of the FATA Weekly Seminar, which takes place every Tuesday from 3:00 - 4:00 PM in Room 422, Sir Alwyn Williams Building and on Zoom https://uofglasgow.zoom.us/j/83611964233?pwd=CgRyzxK8Z9fP2ULTb5ONWZeUYx2t2E.1

Xinghai Hu IR Seminar

Group: Information Retrieval (IR)
Speaker: Xinghai Hu, Tiktok
Date: 09 December, 2024
Time: 15:00 - 16:00
Location: Sir Alwyn Williams Building, 422 Seminar Room

TBC

[FATA Seminar] Bigraphs as a theoretical base for verifying interactive systems

Group: Formal Analysis, Theory and Algorithms (FATA)
Speaker: Cyril ALLIGNOL and Celia PICARD, Ecole Nationale de l'Aviation Civile
Date: 10 December, 2024
Time: 15:00 - 16:00
Location: Sir Alwyn Williams Building, 422 Seminar Room

Abstract: Interactive systems are multiplying, including in critical aeronautical systems such as aircraft cockpits and air traffic control systems. At the same time, UIDLs, the languages that enable the development of such systems, are also on the rise. However, to date, these languages offer no guarantees, neither in terms of compilation nor on the systems developed. Hence the need for a verified programming framework for UIDLs. This is the main subject of our research.

In this talk, we will present the results we have obtained so far and our ongoing work, namely: using bigraphs to express the semantics of UIDLs; defining a formal minimal UIDL based on bigraphs; formalising the bigraph theory in the Coq proof assistant. We will also discuss our future work around the verification of interactive systems, including the verification of interactive properties.

-----------------------------------

This event is part of the FATA Weekly Seminar, which takes place every Tuesday from 3:00 - 4:00 PM in Room 422, Sir Alwyn Williams Building and on Zoom https://uofglasgow.zoom.us/j/83611964233?pwd=CgRyzxK8Z9fP2ULTb5ONWZeUYx2t2E.1

[FATA Seminar] Formally verified hardening of C programs against fault injection

Group: Formal Analysis, Theory and Algorithms (FATA)
Speaker: Basile PESIN, Ecole Nationale de l'Aviation Civile
Date: 10 December, 2024
Time: 15:00 - 16:00
Location: Sir Alwyn Williams Building, 422 Seminar Room

Abstract: Fault attacks allow malicious actors to modify the behavior of a program by
physically injecting a fault in the hardware. They typically target sensitive
applications such as cryptography services, authentication or boot-loader and
firmware updater. They can be defended against by adding countermeasures, that
is control flow checks and redundancies, either in the hardware, or in the
software running on it. In particular, software countermeasures may be added
automatically during compilation.

In this talk, we will describe a formally verified implementation of this
approach in the CompCert verified compiler for the C language. We proposed a
toolkit to implement countermeasures as transformations of a middle-end
representation of CompCert, RTL.  We applied this toolkit to two existing
countermeasures that protect the control flow of the program.  We proved that
these countermeasures are correct, that is, they do not change the observable
behavior of the program during an execution without fault injection. We then
modeled the effect of a fault on the behavior of the program as an extension of
the semantic model of RTL. We used this new model to formally prove the efficacy
of the countermeasure: all attacks are caught. In addition to this formal
reasoning, we evaluated the protected program using Lazart, a tool for symbolic
fault injection.

-----------------------------------

This event is part of the FATA Weekly Seminar, which takes place every Tuesday from 3:00 - 4:00 PM in Room 422, Sir Alwyn Williams Building and on Zoom https://uofglasgow.zoom.us/j/83611964233?pwd=CgRyzxK8Z9fP2ULTb5ONWZeUYx2t2E.1

Yougang Lyu IR Seminar

Group: Information Retrieval (IR)
Speaker: Yougang Lyu, University of Amsterdam
Date: 16 December, 2024
Time: 15:00 - 16:00
Location: Sir Alwyn Williams Building, 422 Seminar Room

TBC

Xiaoyu Zhang IR Seminar

Group: Information Retrieval (IR)
Speaker: Xiaoyu Zhang, Shandong University
Date: 06 January, 2025
Time: 15:00 - 16:00
Location: Sir Alwyn Williams Building, 422 Seminar Room

TBC

Francesco L. De Faveri IR Seminar

Group: Information Retrieval (IR)
Speaker: Francesco L. De Faveri, University of Padua
Date: 13 January, 2025
Time: 15:00 - 16:00
Location: Sir Alwyn Williams Building, 422 Seminar Room

TBC