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 at the following routes.
| Date | Speaker | Title/Abstract |
|---|---|---|
| Monday 2pm 31 Oct Room: 4B96 |
Will Browne (Hosted by John R. Woodward) received a BEng Mechanical Engineering, Honours degree from the University of Bath, UK in 1993, MSc in Energy (1994) and EngD (Engineering Doctorate scheme, 1999) University of Wales, Cardiff. After eight years lecturing in the Department of Cybernetics, University of Reading, UK, he was appointed to School of Engineering and Computer Science, Victoria University of Wellington, NZ in 2008. Associate Professor Browne's main area of research is Applied Cognitive Systems. This includes Learning Classifier Systems, Cognitive Robotics, and Modern Heuristics for industrial application. Blue skies research includes analogues of emotions, abstraction, memories, dissonance and machine consciousness. He is an Associate Editor for Neural Computing and Applications, and Applied Soft Computing. He has published over 100 academic papers, including in IEEE Transactions on Evolutionary Computation on scalable learning and two best paper awards in Genetic and Evolutionary Computation Conference. | Learning Classifier Systems (LCS) Tutorial . This tutorial will introduce the concepts of Learning Classifier Systems (LCS), which have been described as wondrous and inventing, but also as a swamp - this tutorial offers a boardwalk through the swamp. 'Wondrous' as Learning Classifier Systems combine the global search of Evolutionary Algorithms with the local optimisation of Reinforcement Learning to address classification and regression problems. The extracted knowledge though interacting with data or embedded in an environment is human readable. 'Inventing' as LCS' flexible nature allows application to many domains with many types of feedback on solution progress. But 'swampy' as an LCS is not a one line algorithm with separable methods and easily tuned parameters. 40 years of research on LCS has clarified understanding, produced algorithmic descriptions, determined 'sweet spots' for parameters and delivered understandable 'out of the box' code. This tutorial offers a user-friendly guide so that you will be able to proficiently implement LCS. Tutorial participants will have online access to code to work through the examples themselves afterwards. | Monday 4pm 31 Oct Room: LTB3 |
Will Browne (Hosted by John R. Woodward) received a BEng Mechanical Engineering, Honours degree from the University of Bath, UK in 1993, MSc in Energy (1994) and EngD (Engineering Doctorate scheme, 1999) University of Wales, Cardiff. After eight years lecturing in the Department of Cybernetics, University of Reading, UK, he was appointed to School of Engineering and Computer Science, Victoria University of Wellington, NZ in 2008. Associate Professor Browne's main area of research is Applied Cognitive Systems. This includes Learning Classifier Systems, Cognitive Robotics, and Modern Heuristics for industrial application. Blue skies research includes analogues of emotions, abstraction, memories, dissonance and machine consciousness. He is an Associate Editor for Neural Computing and Applications, and Applied Soft Computing. He has published over 100 academic papers, including in IEEE Transactions on Evolutionary Computation on scalable learning and two best paper awards in Genetic and Evolutionary Computation Conference. | Cognitive Learning using Evolutionary Computation Artificial Cognitive Systems encompasses machine intelligence systems, such as robots, that interact with their environment. This talk will highlight research that enables such systems to learn and adapt to problems in their domain and in related domains. The symbolic evolutionary computation technique of Learning Classifier Systems (LCSs) was conceived 40 years ago as an artificial cognitive system. The work presented shows how LCSs can utilise building blocks of knowledge in heuristics ('if-then' rules) to transfer learnt knowledge from small to large scale problems in the same domain. Furthermore, the use of these rules enables functionality learned in sub-problems to be transferred to related problems. Results show that provided the human experimenter can set a rough curriculum for learning concepts, the underlying patterns/models in a problem domain can be learnt in an interpretable manner. | Friday 4th Nov Room: 4B96 |
Wei Pang (Hosted by John R. Woodward) is a Lecturer in Computing Science at the University of Aberdeen. His research interests centre around two strands: (1) bio-inspired computing as applied to data mining and systems identification and (2) qualitative reasoning. Within the first strand, he is particularly interested in artificial immune systems and swarm intelligence, and he has applied these approaches to a wide range of problems, including systems biology, bioinformatics, and data clustering. He was PI on a dot.rural (RCUK Digital Economy Hub) partnership project (£109K, 2014-2015), CoI (sub-project leader) on the ESRC funded social media enhancement project (£541K, 2014-2016, ES/M001628/1), and CoI on a systems biology (multi-scale modelling of biological pathways) project funded by NSFC China (RMB 180K, 2017-2019, Grant No. 61602209). He is a member of the 2015 Scottish Crucible and PI on the follow-up project (2016-2017), a shortlisted candidate for the Royal Society Pairing Scheme (2015 and 2016), and the awardee of the SICSA PECE bursary (2016), Digital Catapult Contributor Membership (2015-2017, valued at £10,000 per annum), British Council RING grant (2012), and FEBS Youth Travel Fund (2009). He has published over 60 papers (including 30 journal papers) in the fields of bio-inspired computing, data mining, systems biology, bioinformatics, and most recently, sociology. URL: http://www.abdn.ac.uk/ncs/people/profiles/pang.wei | Qualitative and Semi-quantitative Modelling for Complex Dynamic Systems . Qualitative and semi-quantitative models are complementary representations to quantitative models. They can provide, at varying precision, a global picture of the behaviour of a system even when the knowledge is incomplete and data are sparse and imprecise. Although in many problems it is difficult to obtain sufficient time series data for numerical system identification, it is still possible to utilise machine learning to identify qualitative models of such systems. In this talk I will present our research on the following three topics: Qualitative Simulation, Semi-quantitative Simulation, and Qualitative Model Learning. In particular, I will talk about immune-inspired approaches to qualitative model learning. Related Software: sites.google.com/site/jmorven/ | Friday 11th Nov Room: 4B96 |
Leticia Hernando (hosted by Gabriela Ochoa) received the BSc degree in Mathematics and the MSc degree in Computer Science from the University of the Basque Country (Spain). She is member of the Intelligent Systems Group at this university since 2010, and she received her Ph.D. degree in Computer Science in 2015. She is currently visiting the Division of Computer Science and Mathematics at the University of Stirling as a postdoctoral researcher. Her research interests include evolutionary computation, permutation-based combinatorial optimisation problems and landscape analysis. | Anatomy of the Attraction Basins: Breaking with the Intuition . Efficiently solving combinatorial optimisation problems requires the development of algorithms that consider the specific properties of the problems. In this sense, local search algorithms are designed over a neighbourhood structure that partially accounts for these properties. Under a neighbourhood, it is the intuition of being in a space resembling the natural landscapes with valleys and mountains. Having this perception, it is commonly believed that, under maximisation, the solutions located in the slopes of the same mountain belong to the same attraction basin, being the peaks of the mountains the local optima. Unfortunately, this is a widespread erroneous visualisation of a combinatorial landscape. Firstly, some results related to the attraction basins suggest being inconsistent with this understanding of the landscape. Secondly, the existence of plateaus commonly found in combinatorial landscapes are usually neglected. This talk provides a detailed analysis of some properties of the attraction basins of instances of permutation-based combinatorial optimisation problems under the 2-exchange and the insert neighbourhoods. Moreover, a way of visualising the attraction basins by means of directed graphs is proposed and different examples are shown. | Friday 18th Nov Room: 2B84 |
Michael Lones (Hosted by John R. Woodward) is an Assistant Professor in the School of Mathematical and Computer Sciences at Heriot-Watt University. He holds MEng and PhD degrees from the University of York, awarded in 1999 and 2003 respectively. Following his PhD, he held an ERCIM Fellowship at the Norwegian University of Science and Technology. He then worked for 8 years as a Research Fellow at the University of York. His research interests span a number of topics at the interface between computer science and biology, including bio-inspired computing, computational biology, and biomedical data mining. He has published over 50 papers in these areas. He is an associate editor of BioSystems, on the editorial board of Genetic Programming and Evolvable Machines, and is a member of the IEEE Technical Committee on Bioinformatics and Bioengineering. | Computing With Cells . Once viewed as mere bags of chemicals, decades of biological enquiry tells us that cells are intricately structured dynamical systems that govern all the activities of biological organisms. In this seminar, I will talk about executable computational models that are inspired by the structure and function of biological cells. I will discuss why we might want to build these, what they can be used for, and how they might complement more conventional neural models in the study of connectionist systems. These ideas will be illustrated using examples from my own work in the areas of robotics, complex systems control and biomedical signal processing. I will also speculate about how unconventional models such as these might be usefully integrated with contemporary approaches in genetic programming, such as hyperheuristics and genetic improvement. | Friday 25th Nov Room: LTA6 |
Dr. Roberto Pagliari (Hosted by Fabio Daolio) is a Data Scientist at ASOS.com, where he works on recommender systems and leads the R&D activities within the Data Science Team. Prior to joining ASOS.com, he was at Qualcomm Inc. (USA) and Vencore Labs (USA). He holds a PhD from Cornell University, and co-authored several publications within the signal processing literature. | State-of-the-art Recommender Systems . This talk is an introduction to state-of-the-art recommender systems. A brief introduction about recommender systems will be followed by a high-level overview of matrix factorisation, which is the machine learning algorithm currently by major internet businesses to provide a better user experience to their customers. The presentation will conclude with some thoughts about current and future trends of recommender systems, such as deep-learning and embedding. | Friday 2nd Dec Room: LTW1 |
Matthew Hubbard (Hosted by Carron Shankland) is currently an Associate Professor in Scientific Computation in the School of Mathematical Sciences at the University of Nottingham where I've been for about 3 years, having moved from the School of Computing at the University of Leeds. | Modelling Drug Delivery and Tumour Growth . Abstract: Predicting the growth of tumours and their response to treatment provides many challenges to mathematical and computational modellers. In this talk I will present two aspects of my recent research in this area: (i) the development of a complex (multiphase) model of tumour growth, and (ii) the parameterisation and application of a simple model of drug transport in tissue. I will focus on how the models are constructed and discuss the challenges in producing the computational simulations which will be presented to demonstrate the model behaviour. | Friday 9th Dec Room: 4B96 |
Dimitrios Milios (Hosted by Andrea Bracciali) is a research associate at the University of Edinburgh and member of the Institute for Adaptive and Neural Computation in the School of Informatics. He received a BSc degree in Computer Science from the Aristotle University of Thessaloniki (Greece), and a MSc degree in Informatics from the University of Edinburgh. He holds a PhD from the University of Edinburgh his thesis has been focused on approximating the stochastic behaviour of Markovian process algebra models. His research interests revolve around combining machine learning and formal methods for the analysis of stochastic systems. | Smoothed Model Checking for Uncertain Continuous-Time Markov Chains . Novel applications of formal modelling such as systems biology have highlighted the need to extend formal analysis techniques to domains with pervasive parametric uncertainty. We consider the problem of computing the satisfaction probability of a formula for stochastic models with parametric uncertainty. We show that this satisfaction probability is a smooth function of the model parameters under mild conditions. This enables us to devise a novel Bayesian statistical algorithm which performs model checking simultaneously for all values of the model parameters from observations of truth values of the formula over individual runs of the model at isolated parameter values. This is achieved by exploiting the smoothness of the satisfaction function: by modelling explicitly correlations through a prior distribution over a space of smooth functions (a Gaussian Process), we can condition on observations at individual parameter values to construct an analytical approximation of the function itself. Extensive experiments on non-trivial case studies show that the approach is accurate and considerably faster than naive parameter exploration with standard statistical model checking methods. | Wednesday (3pm) 14th Dec Room: 4B96 |
Marco Tomassini (hosted by Gabriela Ochoa) is a professor of Computer Science at the Information Systems Department of the University of Lausanne, Switzerland. He got a Doctor's degree in theoretical chemistry from the University of Perugia, Italy, working on computer simulations of condensed matter systems. His research revolves around the structure and dynamics of complex systems. Currently, his main research interests are the modelling of evolutionary games in networks, experimental games, complex networks and systems, and the structure of hard combinatorial search spaces. He has also been active in evolutionary computation, especially spatially structured systems, genetic programming, and the structure of program search spaces. He has been Program Chair of several international events and has published many scientific papers and several authored and edited books in these fields. He has received the EvoStar 2010 Award in recognition for outstanding contribution to evolutionary computation. | Games on networks: models and human behaviour . Evolutionary game theory has been introduced essentially by biologists in the seventies and has immediately diffused into economical and sociological circles. Today, it is a main pillar of the whole edifice of game theory and widely used both in theory and in applications. In this talk, after a gentle introduction to the basics, we shall illustrate the extension of the approach to populations of agents that interact through a network of contacts. Both static and dynamic networks will be considered and their effect on the games’ equilibria will be compared with the usual well mixed populations. We shall also hint at the results of recent experimental work and compare those with the prediction of the theory or of numerical simulations of models. |
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