PGR Seminar – Erdem Kus & Junyu Zhang

You are warmly invited to the next PGR Seminar.

Date & Time: Monday 20/10/2025 14:00-15:00

Location: JC 1.33A

  1. Speaker: Erdem Kus

Title: Frugal Algorithm Selection for Combinatorial Search

Abstract: Solvers for combinatorial search and optimisation problems often exhibit highly complementary performance: instances that are hard for one solver may be easy for another. The Algorithm Selection Problem (ASP) addresses this by predicting, for each problem instance, which solver will perform best. Machine learning models trained for this purpose, however, are typically expensive to construct, as they require exhaustive solver runs on all training instances to obtain ground-truth performance data.

In this work, we propose a frugal alternative that formulates algorithm selection as an active learning problem. Instead of uniformly evaluating all solver–instance pairs, our method intelligently selects the most informative ones, thereby drastically reducing the cost of data collection. We show that standard active learning techniques are inadequate for this setting, as they overlook the structure and cost characteristics unique to algorithm selection. To address this, we introduce novel, cost-aware active learning strategies that leverage auxiliary models to balance informativeness and evaluation cost.

Bio: Erdem is a PhD candidate whose research focuses on Artificial Intelligence (AI) and Constraint Programming (CP).

  1. Speaker: Junyu Zhang

Title: Remaking Characters in Heritage Contexts to Support Inclusive Learning

Abstract: Characters in immersive environments have the potential to enrich user experience, improving engagement with heritage and in so doing benefiting heritage organisations and their communities. Creating authentic digital scenes based upon survey, archaeological and historical data, co-creative design and community engagement enables communities and their visitors to understand the past better. The understanding of authenticity stimulates the potential of enriching cultural heritage with the details of lives past and also discusses how this research benefits the Sustainable Development Goals.

Bio: Minty is a PhD candidate exploring the authenticity of characters to support inclusive learning in heritage contexts. She is interested in how digital technologies can be used in the intersection of different disciplines to achieve SDGs in the field of cultural heritage, so as to enhance the promotion, representation, and well-being in digital humanities education and also affect resonated dialogue and thinking among diverse people and communities in facing the current challenges.

We hope you can join us!

PGR Seminar – Qurat ul ain Shaheen

You are warmly invited to the next PRG Seminar.

Date & Time: Monday 13/10/2025 14:00-14:40

Location: JC 1.33A

Speaker: Qurat ul ain Shaheen

Title: A Framework for Uncertainty Sampling in Active Learning

Abstract: Uncertainty sampling is an active learning paradigm where data instances representing maximum uncertainty for a machine learning model are selected for training. This talk will explore existing uncertainty modelling approaches for binary classification of categorical data.  It will introduce a conceptual framework to improve uncertainty modelling and present some preliminary results.

Bio: Qurat ul ain Shaheen is a final year PhD researcher. Her research focuses on modelling uncertainty in active learning.

We hope you can join us!

PGR Seminar – David Morrison

You are warmly invited to the next PRG Seminar.

Date & Time: Monday 06/10/2025 14:00-14:40

Location: JC 1.33A

Speaker: David Morrison

Title: Synthetic Whole Slide Image Patch Embeddings for Multiple Instance Learning

Abstract: Obtaining high-quality data is a persistent challenge for the training of computational pathology models. As medical data, Whole-slide images (WSIs) are often held under restrictive terms by medical institutions and, as a result, are hard to access by researchers. Where data is available, the number of whole slide images can be limited and skewed towards common pathology types. In addition, there can be issues with labelling: slide-level labels may lack information about specific pathologies, for example, they may be limited to binary labels of normal or malignant, while annotations at the level of patches are rarely available.

Synthetic data generation is a possible solution to these problems by allowing researchers to produce data on demand that can be used in an unrestricted manner with high-quality labels. I have previously presented on the generation of synthetic patch data. In this talk, I will discuss an extension to this work in which this approach is combined with models trained to characterise the slide as a whole in order to provide a synthesis process for data for use with multiple instance learning techniques, commonly used in whole slide image classification.

We hope you can join us!

 

PGR Seminar – Sharon Pisani & Mirza Hossain

The next PGR seminar is taking place this Friday 3rd October 11:00-12:00 in JC 1.33A.

Below are the Titles and Abstracts for Sharon and Mirza’s talks – Please do come along if you are able.

Sharon Pisani

Title: Building Sustainable Heritage Virtual Museums for Communities using Sociodata

Abstract: Virtual museums are moving beyond simple digitisation of artefacts to become dynamic platforms for community engagement and sustainable development. This talk introduces the VERA Platform, which combines a flexible Virtual Museum Infrastructure with a new layer of sustainability-oriented contextual data called sociodata. Sociodata links heritage objects to their cultural landscapes, local communities, and relevant Sustainable Development Goals, enabling richer discovery, analysis, and reuse. In this talk, I will outline the platform’s architecture and metadata model. The talk will highlight technical challenges such as interoperability with European data spaces, and supporting interactive storytelling at scale—issues highly relevant to digital infrastructure and data-driven research in the heritage sector.

Bio: Sharon is a PhD researcher examining the role of emergent digital technologies in preserving and engaging with cultural heritage while supporting sustainable development. Her research focuses on digitising cultural landscapes—both natural and cultural heritage—to assess various impacts on heritage and community identities. She explores how digital tools, including 3D scanning, 3D modeling, and mixed reality, can aid in recreating and safeguarding heritage at risk.

Mirza Hossain

Title: Fishing for monosemantic neurons in histopathology foundation models

Abstract: This early-stage study introduces Histoscope, an interactive system for examining sparse autoencoders (SAEs) that are trained on top of the UNI pathology encoder. Vision transformers for histopathology often exhibit superposition, where single neurons respond to multiple distinct tissue patterns, making interpretation difficult. Histoscope provides quantitative metrics and visualisations to assess whether neurons are monosemantic—associated with a single concept—or polysemantic—associated with multiple concepts. The work highlights methods for analysing internal representations of histopathology foundation models and contributes to efforts toward more transparent AI in pathology.

Bio: Mirza Hossain is a second-year PhD candidate in Computer Science at the University of St Andrews. His research focuses on multimodal AI in medical imaging with an emphasis on mechanistic interpretability of large foundation models. He is supervised by Dr. David Harris-Birtill.

 

Research Activities by Open Virtual Worlds Research Group

Throughout July, the research group Open Virtual Worlds from the School of Computer Science was involved in digitalising the archaeological artefacts, historical sites, and natural landscapes in collaboration with their important partner, Timespan Museum in Helmsdale, Highlands.

Two PhD students, Junyu Zhang and Sharon Pisani, who are researching digital heritage and sustainability, were using 360 photogrammetry to document the local area, which included the Helmsdale Harbour, the former fish curing yards, the Jurassic Coast, the Flow Country UNESCO World Heritage Site, and other natural reserves.

The efforts and work from the Open Virtual Worlds will contribute to the HERITALISE project, which includes seventeen partners from seven different European countries – the Netherlands, Italy, Spain, Belgium, Scotland, Malta, and Cyprus, forming an interdisciplinary group to bring about intelligence, methodology and expertise towards the preservation and promotion of cultural heritage.

old and new harbours in Helmsdale

old and new harbours in Helmsdale

old and new harbours in Helmsdale

old and new harbours in Helmsdale

outdoor fieldwork at natural reserve sites, historical landscapes and heritage ruins

outdoor fieldwork at natural reserve sites, historical landscapes and heritage ruins

members from Open Virtual World and Timespan Museum

members from Open Virtual World and Timespan Museum

members from Open Virtual World and Timespan Museum

Interviews with descendants of historical heritage site

3D scanning of archaeological artefacts

3D scanning of archaeological artefacts

Helmsdale

🎉Award Congratulations to Dr Di Wu – IEEE IoT PhD Thesis Competition 🎉

Congratulations to Dr Di Wu for an outstanding achievement of third place in the inaugural IEEE  IoT PhD thesis competition. Di’s thesis, “Distributed Machine Learning on Edge Computing Systems” truly stood out among the impressive nominations. The judging panel, comprising leading experts in the field, meticulously evaluated each submission based on the following criteria: Problem Definition, Design Methodology, Achievement and Significance, Originality and Innovation, Impact, and Quality of Presentation. Di’s work exemplified excellence across these critical areas.

Congratulations once again to Di on this well-deserved recognition for advancing the field of IoT! 🎉

PhD Student Tai Nguyen Wins Best Paper Award at GECCO 2025

Tai Nguyen, supervised by Dr. Nguyen Dang (University of St Andrews) and Dr. Carola Doerr (CNRS Researcher, Sorbonne Université, France), has been awarded the prestigious Best Paper Award at GECCO 2025, held in Málaga, Spain, from July 14–18, 2025. GECCO (Genetic and Evolutionary Computation Conference) is the leading international conference in its field, known for its highly competitive acceptance rate (36% in 2025).
The award-winning paper is titled: “On the Importance of Reward Design in Reinforcement Learning-based Dynamic Algorithm Configuration: A Case Study on OneMax with (1+(λ,λ))-GA.” link: https://dl.acm.org/doi/10.1145/3712256.3726395

To help a general audience understand the significance of this work, here is a brief and simplified summary:

This research tackles a central challenge in Artificial Intelligence (AI) and optimization. Imagine trying to complete a complex task by finding the best possible strategy, which often involves tuning many parameters—a process known as “algorithm configuration.” Now, imagine an intelligent AI assistant that not only helps determine the best settings for the task at hand but also adapts those settings as the task evolves. This is known as “Dynamic Algorithm Configuration” (DAC). Tai’s work applies Reinforcement Learning (RL), a powerful AI technique in which computers learn to make decisions by receiving “rewards” or “penalties” based on their actions. The paper’s key insight is that the design of the reward system plays a crucial role in enabling the AI assistant to learn effectively, particularly in dynamic or complex environments. If rewards are not carefully crafted, the AI may struggle to identify effective strategies, especially as problem sizes increase. To address this, the team developed methods for designing reward functions, which significantly improved the AI’s learning speed and performance. In essence, their approach allows the AI to find optimal solutions much more efficiently than existing techniques.

Congratulations to Tai Nguyen, Dr. Nguyen Dang, Dr. Phong Le (University of St Andrews), Dr. André Biedenkapp (University of Freiburg, Germany), and Dr. Carola Doerr (Sorbonne Université, France) on this outstanding achievement! Party popper