Seminar – Richard Connor – 5th November

The second school seminar on 5th November at 2pm, on Teams.  If you do not have the Teams link available please contact the organiser, Ian Gent.

Dimensionality Reduction in non-Euclidean Spaces
Richard Connor
Deep Learning (ie Convolutional Neural Networks) gives astoundingly good classification over many domains, notably images. Less well known, but perhaps more exciting, are similarity models that can be applied to their inner layers, where there lurk data representations that can give a much more generic notion of similarity. The problem is that these data representations are huge, and so searching a very large space for similar objects is inherently intractable.
If we treat the data as high-dimensional vectors in Euclidean space, then a wealth of approximation techniques is available, most notably dimensionality reduction which can give much smaller forms of the data within acceptable error bounds. However, this data is not inherently a Euclidean space, and there are better ways of measuring similarity using more sophisticated metrics.
The problem now is that existing dimensionality reduction techniques perform analysis over the coordinate space to achieve the size reduction. The more sophisticated metrics give only relative distances and are not amenable to analysis of the coordinates. In this talk, we show a novel technique which uses only the distances among whole objects to achieve a mapping into a low dimensional Euclidean space. As well as being applicable to non-Euclidean metrics, its performance over Euclidean spaces themselves is also interesting.
This is work in progress; anyone interested is more than welcome to collaborate!

Learning to Describe: A New Approach to Computer Vision Based Ancient Coin Analysis

The work on deep learning based understanding of ancient coins by Jessica Cooper, who is a Research Assistant and a part-time PhD student supervised by Oggie Arandjelovic and David Harrison has been chosen as a featured, “title story” article by the Journal Sci where it was published in a Special Issue Machine Learning and Vision for Cultural Heritage.

Zoë Nengite awarded Principal’s Medal

Congratulations to Zoë Nengite who has been awarded The Principal’s Medal in recognition of outstanding academic achievement and exceptional activities within the University and the wider St Andrews community. The Medal is awarded to students who have both excellent academic accomplishments and those who have inspired and supported their peers and who have often undertaken extensive advocacy work, which has improved life for many of their fellow students.

Zoë sent us a reflection on time spent studying in the School and a photo celebrating with Mum.

“I’m really sad that my time at St Andrews has come to an end. I will especially miss the School of Computer Science. We are such a close community of students and staff alike. I will even miss the Jack Cole labs, despite spending many hours with my head in my hands stuck on a problem gripping my mug of coffee. I always knew that help wasn’t too hard to find.

“Some of my best memories are from my time at St Andrews. Most of them spent with my closest friends who also studied Computer Science. Coming from London, I was apprehensive about St Andrews, but it quickly became a place I called home. I think even years from now, it will always be somewhere I call home.”

The award was announced during the virtual conferral of degrees in July. Zoë hopes to attend a rescheduled Class of 2020 Ceremony in the future where we look forward to celebrating with her in person.

Leverhulme Early Career Fellowship for Nguyen Dang

Congratulations to Dr Nguyen Dang, who has been awarded a Leverhulme Trust Early Career Fellowship. The 3 year Fellowships are intended to assist those at an early stage of their academic careers to undertake a significant piece of publishable work. Nguyen will be researching Constraint-based automated generation of synthetic benchmark instances.

Abstract summary: “Combinatorial problems such as routing or timetabling are ubiquitous in society, industry, and academia. In the quest to develop algorithms to solve these problems effectively, we need benchmark instances. An instance is an example of the problems at hand for testing how well an algorithm performs. Having rich benchmarks of instances is essential for algorithm developers to gain understanding about the strengths and weaknesses of their approaches, and ensure successful applications in practice. This fellowship will provide a fully automated system for generating valid and useful synthetic benchmark instances based on a constraint modelling pipeline that supports several algorithmic techniques.”

Winnability of Klondike Solitaire research features in Major Nelson’s video podcast

Research carried out by Charlie Blake and Ian Gent to compute the approximate odds of winning any version of solitaire features in Major Nelson’s Video Podcast [Interview with Ian and Charlie starts 23:56] for XBox news today.

Today is National Solitaire Day and the 30th anniversary of the game. The celebrations include an invitation to participate in a record breaking attempt at the most games of Microsoft Solitaire completed in one day. You can download the collection free or play it through your browser.

The Klondike Solitaire research also featured in the New Scientist last year.
Link to the full paper on arxiv: https://arxiv.org/abs/1906.12314

Online article published in Technology Nov 17th 2019: https://www.newscientist.com/article/2223643-we-finally-know-the-odds-of-winning-a-game-of-solitaire/