Accessibility statement

2018/19 seminar archive

Autumn 2018

10 October: A Computational Approach to Historical Linguistics
Giuseppe Longobardi (Lang.&Ling.Sci, UoY) and Dimitar Kazakov (CS, UoY)
Panopto recording: [link]

17 October: On the limits of AI/Deep Learning
Mark Bishop (Comp., Goldsmiths)
Panopto recording: [link]

24 October: Model Integration and Decision-Making for Self-Adaptation in Mobile Robotic
Javier Cámara Moreno (CS, UoY)
Panopto recording: [link]

31 October: Semantic technologies in practice: Designing, building, maintaining and exploiting knowledge graphs at Ontotext
Nikola Tulechki (Ontotext)
Panopto recording: [link]
Online slides: [link]

7 November: Medical Imaging Informatics: An opportunity to revolutionize health care by working together with the medical professionals
Bishesh Khanal
Panopto recording: [link]

14 November: Supporting dynamic reconfigurations in flexible manufacturing
Nikolay Mehandjiev (Alliance Manchester Business School, U. of Manchester)

28 November: Modelling movement of robot swarms and other herds
Vashti Galpin (School of Informatics, U. of Edinburgh)

Spring 2019

30 January: Healthcare robotics – from intelligent design to AI
Sanja Dogramadzi (University of the West of England, Bristol Robotics Lab)

Abstract: This presentation gives an overview of healthcare robotics portfolio at Bristol Robotics Laboratory that spans from surgical to assisted living robots as well as robotic systems to support radiotherapy. The fast and widespread use of robotic technologies in healthcare applications aims at improving patient outcomes and reducing NHS costs. The robotic technologies include soft sensors and actuators, complex robotic system design and control, haptics, machine learning and sensor processing. 

6 February: Playful technology and persuasion
Jo Iacovides (CS, UoY)

Abstract: When people talk about games and play the first thing that comes to mind is normally “fun”.However, by focusing primarily on positive emotional experiences, are we missing out on potentialopportunities to elicit more in-depth engagement with a particular issue or domain? In this talk, I willdiscuss three approaches to developing playful technologies that aim to persuade people to thinkand act in different ways. The first emphasises novelty and curiosity, where a physical computinginstallation, the Green Squeezy Balls kiosk, was located in a public space to raise awareness andinvite discussion on environmental actions. The second involves the development of serious gamesfor safety-critical contexts, including law enforcement and healthcare, which simulate aspects ofreal-world practice for the purpose of training and education. The third focuses on more negativeforms of experience through playing games that involve a level of discomfort. These games weredeveloped as part of a persuasive game design competition linked to a research project on medicaldevice safety; the competition was a public engagement activity which aimed to prompt reflectionon the themes of human error and blame culture in context of healthcare. I will discuss the extent towhich each of these approaches was successful and highlight particular issues that need to beconsidered when attempting to use play and games for different persuasive purposes.

18 February: BBC's Collectaholics
Jim Austin (CS, UoY) live and on TV

20 February: Virtualised Queueing Theory
Florin Ciucu (CS, Warwick)

Abstract: Queueing is a fundamental problem characteristic toresource-sharing systems: whenever there is more load thanavailable resources, some jobs/requests are queued and implicitly delayed. The classical queueing theory was conceived by Erlan more than a century ago, and has since evolved as one the most important branches of applied probability. Due to some key technical limitations in its scope, the applicability of queueing theory has been increasingly questioned in the context of complex modern systems, e.g., the Internet.

This talk introduces Virtualized Queueing Theory (VQT), a modern approach whose central idea is to increase tractability by sacrificing the exact analysis. At the apparent expense of computing queueing performance metrics in terms of bounds, VQT can address, for the first time, two fundamental problems: per-flow scheduling and multi-queue networks, for a broader class of arrival processes (e.g., Markov modulated). Moreover, by relying on martingale representations of certain transforms of the queueing systems, the obtained bounds are tight and improve existing results by orders of magnitude.

27 February: Automated Concept Trending in Social Media Using Association Rules
Rayner Alfred (Universiti Malaysia Sabah, Kota Kinabalu, Malaysia)

Topic trending is a popular research topic in recent years since, there are massive participations in social web sites, countless number of updates, news, opinions and product reviews are being constantly posted every day. The identification of popular topics discussed or posted on social media platforms is becoming more important as the new knowledge can be extracted from these findings. In this work, a novel method is proposed to extract popular topics from social media and determine the topic trending based on timeline. The approach uses Twitter’s tweets as the dataset. The data is then pre-processed by undergoing several processes that include stop words removal, stemming, tokenization and Term Frequency-Inverse Document Frequency (TF-IDF) weighting. The k- Means clustering is then performed to cluster each data that consists of processed keywords and collected every day. The popular topics will be then extracted from the clusters and the topic trends will be determined based on the observed frequent patterns and correlation between keywords by using the association rules. The performance of the proposed method is evaluated based on the similarity of the results with the current trends obtained from the Twitter site. The result from the findings shows that the proposed method is able to produce more enriched trends that are similar to current initial trends.

13 March: Using Long Range Wide Area Networks (LoRAWAN) in the real world
Ian Riley (Technical Director) and Chris Brown (Senior Developer), Labman Automation

Abstract: LoRaWAN offers long range, low power consumption data network for IoT applications. Its networking protocol is designed to wirelessly connect  battery operated ‘things’ to the internet in regional, national or global networks, and targets key IoT requirements such as bi-directional communication, end-to-end security, mobility and localization services.  The technology can be utilized by public, private or hybrid networks on the license-free radio frequency bands such as 868MHz in the UK. We will talk over a current project to provision tens of thousands of LoRa enabled devices across multiple networks all over the globe with a relatively high message frequency. We will discuss why we chose LoRa, pitfalls and results to date. We will also go into the implications limited by the LoRa technology, fair usage, bandwidth, environment and infrastructure. We will detail our solution for scaling the solution to a commercial level with redundancy and possible future avenues to increase the effective density of devices within a small geographic location.

Summer 2019

15 May: The Future of Digitalization — Design Trade-offs between Artificial Intelligence and Human-Centered Design
Gerhard Fischer, University of Colorado at Boulder

Abstract: The Future of Digitalization is not out there to be discovered — it will be designed. In design, trade-offs are universal because there are no best solutions independent of goals, objectives, and values, specifically for systemic, ill-defined, and wicked problems such as  Artificial Intelligence and Human-Centered Design. Grounded in research activities from a broad spectrum of different disciplines and an analysis from our research over the last two decades,  the presentation will critically analyze the current hype about Artificial Intelligence by contrasting it with the objectives pursued by Human-Centered Design.

22 May
Marian Ursu, Department of Theatre, Film, Television and Interactive Media, Digital Creativity Labs, UoY

19 June: A Quick Tour of Three Robotics Projects
Victoria Hodge, YorRobots

The data-privacy dilemma: How fully homomorphic encryption could bring healthcare into the digital era
Delaram Kahrobaei (CS, UoY)

Abstract: Imagine this: The key to better cancer treatments is within reach, based on patterns from data that is scattered across various locations all over the world. This data could be digitalised, labelled, collected, stored and interpreted. However, this data belongs to a countless number of individuals – and their right to data privacy weighs just as much as the dream of curing a lethal disease. That is exactly the dilemma we face. Potentially life-saving and cost-cutting solutions, such as significant improvement of early diagnosis, real time remote diagnosis or identification of promising treatment options are impossible because data privacy needs to be ensured. Standard ways of anonymisation cannot reliably prevent re-identification of individual patients. But is there any way in which an algorithm trained on encrypted data could produce informative, actionable results? In a collaboration between computer scientists, mathematicians and computational medical scientists at the City University of New York and University of Michigan (Ann Arbor), we applied fully homomorphic encryption (FHE) to classify breast cancer data as benign or malignant – completely preserving data-privacy. The lesson is clear: FHE could be the foundation for bringing healthcare into the digital era.