Skip to main page content - your browser does not fully support our CSS, or is text-only.

Computing Science and Maths Seminars, 2018/2019

Spring 16 image

Seminars will take place in Room 4B96,  Cottrell Building, University of Stirling. Normally, from 15.00 to 16.00 on Friday afternoons during semester time, unless otherwise stated. For instructions on how to get to the University, please look here.

If you would like to give a seminar to the department in future or if you need more information,  
please contact the seminar organisers, Dr. Sandy Brownlee ( and Dr. Wen-shin Lee (

Autumn 2018

Date Speaker Title/Abstract
7th September
Ken Reid, CSM, University of Stirling Shift Scheduling and Employee Rostering: An Evolutionary Ruin and Stochastic Recreate Solution
For decades, since the inception of the field, scheduling problems have been solved with a variety of techniques. Many proven algorithms to these problems exist; however, there is no single method to solve all the vast variety of problems that exist across many sub-fields with differing datasets. In this paper we explore the use of an Evolutionary Ruin & Stochastic Recreate algorithm, with a Simulated Annealing control mechanism, to a real-world employee scheduling problem and its ability to solve this problem to near optimality. The combinatorial possibilities of parameterisation are very large - the Taguchi design of experiments method is used to examine a subset of those possibilities within a limited runtime budget. Evolutionary Ruin and Stochastic Recreate has not previously been applied to the specific scheduling domain of employee scheduling and rostering: we investigate the effectiveness of the algorithm with different parameter values and discuss the insight it provides into the runtime effect of the mechanisms of Evolutionary Ruin & Stochastic Recreate.Â
14th September
Prof. Mark Giesbrecht, David R. Cheriton School of Computer Science, University of Waterloo, Canada Eigenvalues, invariant factors and algorithms for sparse integer matrices
Integer matrices are often characterized by the lattice of combinations of their rows or columns. This is captured nicely by the Smith canonical form, a diagonal matrix of invariant factors, to which any integer matrix can be transformed through left and right multiplication by unimodular matrices. Algorithms for computing Smith forms have seen dramatic improvements over the past 40 years, but effective algorithms for large sparse matrices still need improvement.
Integer matrices also possess complex eigenvalues and eigenvectors, and every such matrix is similar to a unique one in Jordan canonical form. There is a wealth of numerical methods for computing eigenvalues, and Krylov-type algorithms are effective for sparse matrices.
It would seem a priori that the invariant factors and the eigenvalues would have little to do with each other. Yet we will show that for “almost all” matrices the invariant factors and the eigenvalues are equivalent under a p-adic valuation, in a very precisely counted sense.
A much-hoped-for link is then explored for fast computation of Smith forms of sparse integer matrices, via the better understood algorithms for computing eigenvalues and effective preconditioning.
All the methods are elementary and no particular background beyond linear algebra will be assumed.
This is joint work with graduate student Mustafa Elsheikh.
Speaker Bio: Mark Giesbrecht is Director of the David R. Cheriton School of Computer Science at the University of Waterloo, where he has been a Professor of CS for the past 16 years. Prior to this he was an Assistant Professor at the University of Manitoba and the University of Western Ontario. His research is in symbolic computation and computer algebra, for which he was named an ACM Distinguished Scientist and was a co-winner of the NSERC Synergy Award in 2004. He was the Chair of ACM SIGSAM and has chaired the steering committee for ACM ISSAC (the premiere computer algebra conference) as well as serving as program chair and program committee member.
21st September
Dr. Anna Kirpichnikova, CSM, University of Stirling Focusing wave in unknown media
We consider a combination of two powerful mathematical methods: the time-reversal and the boundary control methods. The first one allows to find the scatterer inside the domain by "sending the wave back in time" from the boundary, the second method allows to reconstruct the unknown density inside the body, i.e., the density distribution, from boundary measurements. Combined together, the two methods give a computationally cheap procedure that allows to explore the inside of the body "on the go" and produce a wave that will concentrate near the point of interest inside the body.
Speaker Bio: Anya Kirpichnikova received her BSc and MSc in Mathematics from the St-Petersburg State University, Russia (Department of Applied Mathematics and Control Processes) in 2000. Her MSc Dissertation, co-supervised by mathematicians from the St-Petersburg Department of Steklov’s Mathematical Institute, was focused on the approximation of wave behaviour in the diffraction theory applied to the object with curvature jump at the boundary.
She then moved to the UK in 2001 and obtained her PhD from the Loughborough University in 2005. Her research was concentrated on inverse problems for boundary value problems on complex bodies. She also continued her collaboration with the Steklov’s Institute, investigating wave behaviour for various complex matrices (interfaces, layers, insertions, etc.) in acoustics, elasticity, as well as electromagnetic media.
In 2006 she received the EPSRC Postdoctoral Fellowship and started independent research at the University of Edinburgh expanding her research fields to control and optimisation problems, with the aim to produce computational scheme for applications in medicine for non-invasive tumour testing and improved tumour destruction protocols. In 2010 she moved to the University of Glasgow, where she worked for 3 years as a Lecturer in Mathematics. She joined the Department of Mathematics and Computer Science at the Liverpool Hope University in September 2013.
She is a member of the Organising and Scientific Committee, and also a member of the Editorial Board of the Proceedings of the Annual International Conference "Days on Diffraction" held in St-Petersburg, Russia.​
28th September
Prof. Yvonne Ou, Department of Mathematical Sciences, University of Delaware, USA Interplay between computational mathematics, sciences and other branches of mathematics
The talk will start with a brief introduction of US NSF's 10 Big Ideas and give an overview of the recent trend of computational mathematics in the US. Then I will present my own research topic which will serve as an example to demonstrate how a problem in composite materials can be studied from the computational mathematics point of view. Furthermore, it will be shown how its link to the function theory can help us develop a mathematically simple and elegant way for solving a complicated problem arising in wave propagation of poroelastoc materials.
Speaker Bio: Yvonne Ou received her B.Sc. in Atmospheric Sciences from National Taiwan University in 1993, works as a research assistant in the Mesoscale Lab of NTU from 1993 till 1995 and obtained her Ph.D. in Applied Mathematics from University of Delaware in 2001. Her thesis was on inverse scattering and homogenization. She was a postdoctoral fellow in Institute of Mathematics and its Applications (IMA), University of Minnesota, where she studied the regularity of Navier Stokes equations and started her research in inverse-homogenization of composite materials. Prior to joining the faculty in the Department of Mathematical Sciences at the University of Delaware in 2011, she was a research scientist in the Division of Computational Mathematics in the Oak Ridge National Laboratory, where she worked with computational chemists and nuclear physicists. Currently, she is on leave from University Delaware and serve as a program director of the Computational Mathematics Program in the US National Science Foundation.
5th October
Dr. Misha Feigin, School of Mathematics and Statistics, University of Glasgow (Angular) Calogero-Moser systems and related algebraic structures
Calogero-Moser system is a well-known and important integrable system due to its numerous connections to other areas of mathematics. I am going to overview this system as well as its generalisation for an arbitrary Coxeter group stressing integrability structure, especially Dunkl operators. I’ll also discuss properties of the angular part of the (generalised) Calogero-Moser systems, which is naturally related to Dunkl angular momenta algebra. I’ll also discuss a relation with a version of Laplace-Runge-Lenz vector generalising such a vector for the usual Coulomb problem. The talk is partly based on joint works with T. Hakobyan and A. Nersessian.
Speaker Bio: Misha Feigin graduated from Moscow State University named after Lomonosov in 1997. He received a candidate of science degree from Moscow State University in 2001 and a PhD from Loughborough University in 2003. He worked as Chapman Fellow at Imperial College London during 2003-2005 before taking up a permanent lectureship at University of Glasgow where he is now a Senior Lecturer. Research interests of Misha lie in the are of integrable systems with intersections with Algebra, Geometry and Mathematical Physics. In particular, finite groups generated by reflections often show up in his work.
10th October
Note different
day + time!
Dr. Ahsan Adeel, CSM, University of Stirling Role of Awareness and Universal Context in a Spiking Conscious Neural Network (SCNN): A New Perspective and Future Directions
Awareness plays a major role in human cognition and adaptive behaviour, though mechanisms involved remain unknown. Awareness is not an objectively established fact, therefore, despite extensive research, scientists have not been able to fully interpret its contribution in multisensory integration and precise neural firing, hence, questions remain: (1) How the biological neuron integrates the incoming multisensory signals with respect to different situations? (2) How are the roles of incoming multisensory signals defined (selective amplification or attenuation) that help neuron(s) to originate a precise neural firing complying with the anticipated behavioural-constraint of the environment? (3) How are the external environment and anticipated behaviour integrated? Recently, scientists have exploited deep learning architectures to integrate multimodal cues and capture context-dependent meanings. Yet, these methods suffer from imprecise behavioural representation due to limited contextual exploitation with no integration of overall knowledge of the problem (awareness) at the neural level. In addition, such network level approaches can't be used for deep analysis and information decomposition to understand the neural circuitry and underlying information processing mechanisms with respect to the outside world and anticipated behaviour. In this research, we introduce a novel theory on the role of awareness and universal context that can help answering the aforementioned crucial neuroscience questions. Specifically, we propose a class of spiking conscious neuron in which the output depends on three functionally distinctive integrated input variables: receptive field (RF), local contextual field (LCF), and universal contextual field (UCF) - a newly proposed dimension. The RF defines the incoming sensory signal, LCF defines the information coming from other parts of the brain to support ambiguous RF, and UCF defines the awareness. It is believed that the conscious neuron inherently contains enough knowledge about the situation in which the problem is to be solved based on past learning and reasoning and defines the precise role of incoming multimodal signals to originate a precise neural firing. It is shown that, when implemented within a SCNN, the conscious neuron helps modelling a more precise human behaviour as compared to state-of-the-art unimodal and multimodal models. The SCNN, when exploited to model human's audio-visual (AV) speech processing, performed comparably to deep long-short-term memory (LSTM) network. We believe that the proposed theory could help addressing a range of real-world problems including elusive neural disruptions, human-like computing, sentiment analysis, financial modelling etc.
19th October
No seminar Internal events
26th October
No seminar Mid-semester reading week
31st October
Note different day!
Still in 4B96
Dr. Katherine M. Malan, Department of Decision Sciences, University of South Africa Landscape-aware Constraint Handling Applied to Differential Evolution
In real-world contexts optimisation problems frequently have constraints. Evolutionary algorithms do not naturally handle constrained spaces, so require constraint handling techniques to modify the search process. Based on the thesis that different constraint handling approaches are suited to different problem types, this study shows that the features of the problem can provide guidance in choosing appropriate constraint handling techniques for differential evolution. High-level algorithm selection rules are derived through data mining based on a training set of problems on which landscape analysis is performed through sampling. On a set of diff erent test problems, these rules are used to switch between constraint handling techniques during differential evolution search using on-line analysis of landscape features. The proposed landscape-aware switching approach is shown to out-perform the constituent constraint-handling approaches, illustrating that there is value in monitoring the landscape during search and switching to appropriate techniques depending on the problem characteristics. Results are also provided that show that the approach is fairly insensitive to parameter changes.
2nd November
Dr. Giuseppe di Fatta, Computer Science, University of Reading The Epidemic Paradigm for Decentralised Communication and Computing
Communication and computing in large-scale networked systems typically benefit and often require scalable, decentralised and fault-tolerant approaches. Epidemic protocols provide an interesting paradigm that adopts a randomised communication strategy inspired by the exponential outbreak model of infectious diseases. Their advantages over global communication schemes based on deterministic overlay networks are their inherent robustness, scalability and full decentralisation. They have been shown to be suitable for distributed and dynamic systems of large and extreme-scale. They have been proposed for fundamental services such as information dissemination and data aggregation and for other more complex applications such as distributed data mining, exascale high performance computing and decentralised consensus. The talk provides an introduction to Epidemic protocols and overview of their potential applications in distributed systems.
Short Bio: Dr. Giuseppe Di Fatta is an Associate Professor of Computer Science and the Head of the Department of Computer Science at the University of Reading, UK. In 1999, he was a research fellow at the International Computer Science Institute (ICSI), Berkeley, CA, USA. From 2000 to 2004, he was with the High-Performance Computing and Networking Institute of the National Research Council, Italy. From 2004 to 2006, he was with the University of Konstanz, Germany. His research interests include data mining algorithms, distributed and parallel computing, and data-driven multidisciplinary applications. He has published over 100 articles in peer-reviewed conferences and journals. He has served in the editorial board of the Elsevier Journal of Network and Computer Applications, is the co-founder of the IEEE ICDM Workshop on Data Mining in Networks and has chaired several international events, such as the International Conference on Internet and Distributed Computing Systems.
9th November
4:30pm, in V1
Note different
room and time!
Edinburgh Mathematical Society public lecture: Professor T Zastawniak, University of York
Branching processes representation of solutions to non-linear Dirac equations

For the solutions of the non-linear Dirac equation in two space-time dimensions, we present a Feynman-Kac type representation in terms of branching Poisson processes. These resemble McKean's (1975) branching diffusion representation for the solution to the non-linear reaction- diffusion equation of Kolmogorov-Petrovskii-Piskunov (the KPP equation). Such representations can be established for a wide range of polynomial non-linearities in the Dirac equation, including a number of cases studied in physics such as the Thirring model, the Gross-Neveu model, as well as the non- linear Dirac equations in the context of Feshbach resonance for Bose-Einstein condensates, in periodic dielectric materials under Bragg resonance, or in modelling wave resonances in photonic crystals.
16th November
Dr. Benjamin Lacroix, Computing Science and Digital Media, Robert Gordon University, Aberdeen Optimisation under uncertainty
Most of the academic research effort in optimisation studies deterministic problems. However, models used for real world problems are often based on uncertain variables described by random variables. This can either be imprecise measurements, varying environmental conditions or the occurrence of random events. As a consequence, the fitness of a solution is defined by a posterior distribution of fitness instead of a single value. Estimating this distribution can be challenging and often computationally expensive. Hence, when tackling stochastic optimisation problems, important decision needs to be made on the computational effort dedicated to the evaluation of each solution (number of simulation). A popular example of stochastic problems is maintenance schedule optimisation. The evaluation of maintenance schedule is based on the simulation of the life cycle of a set of industrial assets subject to the random occurrences of failures.

In this talk, we will introduce general concepts of optimisation under uncertainty using the case of a maintenance schedule problem and present the application of racing as selection method in a simple evolutionary framework.
21st November
Note different day
Assoc. Prof. Ansgar Scherp, CSM, University of Stirling About Extreme Analyses of Texts and Graphs
The talk provides an overview about my research in extreme text and data mining. In example, it considers the task of eXtreme Multi-Label Classification (XMLC) on very large document corpora and making recommendations such as scientific papers and citations. Subsequently, I focus on mining Open Data on the Web using a stream-based approach that can operate over extremely large graphs, i.e., billions of edges. A grant challenge is to deal with the evolution of the graph data. To this end, I have investigated the dynamics of entities on the web in order to find, e.g., periodicities in changes etc., and to use this information for data caching, predict future changes, etc. The talk will conclude with examples and motivations why to combine the analyses of text with the evolution of graphs.
23rd November
Winter Graduations
30th November
Dr Deepayan Bhowmik, CSM, University of Stirling Multimedia signal processing for security, surveillance and trust
In this seminar, I'll talk about overview of my current and past research on multimedia signal processing with application to security, surveillance and trust. The presentation will particularly focus on three different areas: 1) image and video processing for surveillance and security, 2) natural language processing for fake news detection on social networks and 3) low power and accelerated embedded vision system design on FPGAs. I'll also talk about a new multimedia blockchain framework that envisage to solve some of the long standing issues around digital rights management in the entertainment industry. Generic understanding of media piracy and related issues can be found from my Innovate UK blog:
Bio: Dr Deepayan Bhowmik is a Lecturer in Computing Science and Mathematics at the University of Stirling. He holds a Ph.D. (2011) in Electronic and Electrical Engineering from the University of Sheffield, UK. His PhD was funded by EPSRC through a prestigious Dorothy Hodgkin Postgraduate Award. He was a research associate at the University of Sheffield, in Sheffield, UK (2011-13) and in the Institute of Sensors, Signals and Systems at Heriot-Watt University, Edinburgh, UK (2013-16). Between 2016 and 2018, he was a lecturer in Computing at Sheffield Hallam University, Sheffield, UK. He received external funding from EPSRC and industries to conduct his research.

Dr Bhowmik is conducting research in signal and image processing for security, surveillance and trust. His core research evolves around adaptive signal/image decomposition and model of computations for signal processing system design. Application domains include multimedia security and forensics, sentiment aware social networks analysis, media blockchain, computer vision based crowd analysis and low power vision systems design on FPGAs. He is part of the editorial board for Elsevier Data in Brief, a guest editor of Elsevier Journal of Information Security and Applications and an active reviewer for major IEEE/IET/ACM/Elsevier journals. He is/was an organising chair of IEEE ICASSP 2019 and IEEE DSP 2017 for special sessions on multimedia cybersecurity.
7th December
Dr. Yunhyong Kim, School of Humanities, University of Glasgow Fostering catalysts for knowledge discovery in a data driven society
This seminar proposes a new conceptual framework for understanding artificial intelligence, digital preservation, and digital forensics as catalysts in a bigger information processing machinery. The seminar presents a comparative examination of historical developments in the three disciplines and their relevance to epistemology. The discussion is then taken further with concrete examples of cross-disciplinary and independent developments, to explore possible mutual benefits, especially in the context of recent discussions regarding AI and ethics. Finally the seminar uses bio-inspired notion of catalysts to highlight how each of the disciplines help meliorate information processing. This work is the first to analyse the relationship between the three disciplines, reflecting three perspectives in parallel: historical, epistemological, and biological.
[an error occurred while processing this directive]

Top image: Illustrated example of running the Epsilon-constraint algorithm in order to maximise two objectives: find an optimal solution for objective 1; restrict the solution space according to the solution's value for objective 2 and look for an optimum solution of objective 1 in that space; repeat the previous step until there are no more solutions to be found. Any dominated solutions need to be filtered out of the set of solutions.
Courtesy of Dr. Nadarajen Veerapen. Related to a recent publication:

N. Veerapen, G. Ochoa, M. Harman and E. K. Burke. An Integer Linear Programming approach to the single and bi-objective Next Release Problem. Information and Software Technology, Volume 65, September 2015, Pages 1-13, ISSN 0950-5849. DOI:10.1016/j.infsof.2015.03.008

This page is maintained by:
Computing Science and Mathematics
Faculty of Natural Sciences
University of Stirling, Stirling FK9 4LA
Tel: +44 1786 46 7421

© University of Stirling FK9 4LA Scotland UK • Telephone +44 1786 473171 • Scottish Charity No SC011159
Portal Logon

Forgotten login?