Computing Science Seminars, Spring 2017
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.
Spring 2017
| Date | Speaker | Title/Abstract |
|---|---|---|
| Friday 20th Jan Room: 2B87 |
Dr David Manlove (Hosted by John Woodward) is a Senior Lecturer in Computing Science at the University of Glasgow, where he has been since 1995. He is interested in algorithms for problems involving matching agents to commodities (e.g., junior doctors to hospitals, kidney patients to donors) in the presence of ordinal preferences or cardinal utilities. He has written or co-authored over 60 papers in this area, and his book "Algorithmics of Matching Under Preferences" was published in 2013. Much of this research has involved designing algorithms to cope with NP-hard optimization problems arising in healthcare-related settings, working in collaboration with the National Health Service in the UK. He is Vice-Chair of the ENCKEP COST Action (European Network for Collaboration on Kidney Exchange Programmes). | Optimizing preferences and social welfare in healthcare-related matching problems . Matching problems typically involve assigning agents to commodities, possibly on the basis of ordinal preferences or other metrics. These problems have large-scale applications to centralised matching schemes in many countries and contexts. In this talk I will describe the matching problems featuring in two such schemes in the UK that have involved collaborations between the National Health Service and the University of Glasgow. One of these dealt with the allocation of junior doctors to Scottish hospitals (1999-2012), and the other is concerned with finding kidney exchanges among incompatible donor-patient pairs across the UK (2007-date). In each case I will describe the applications, present the underlying (NP-hard) algorithmic problems, outline the various solution techniques and give an overview of results arising from real data connected with the matching schemes in recent years. | Friday 27th Jan Room: 4B96 |
Wei Chen (Hosted by John Woodward) is a Research Associate in School of Informatics at University of Edinburgh. He received his PhD from University of Nottingham on Type Theory supervised by Prof. Roland C. Backhouse. In 2012 Wei worked with Prof. Martin Hofmann on type-based verification in Munich. He started his current RA with Prof. David Aspinall since 2013, focusing on learning policies for mobile security. Wei's main research interests are in formal methods, in particular, type theory, combinatorial games, and Buechi automata with their applications in program analysis and verification. He is currently working on combining formal methods and machine learning to help with mobile security. | More Semantics More Robust: Improving Android Malware Classifiers . Automatic malware classifiers often perform badly on the detection of new malware, i.e., their robustness is poor. We study the machine-learning-based mobile malware classifiers and reveal one reason: the input features used by these classifiers can?t capture general behavioural patterns of malware instances. We extract the best-performing syntax- based features like permissions and API calls, and some semantics-based features like happen-befores and unwanted behaviours, and train classifiers using popular supervised and semi-supervised learning methods. By comparing their classification performance on industrial datasets collected across several years, we demonstrate that using semantics- based features can dramatically improve robustness of malware classifiers. | Friday 3rd Feb Room: 4B96 |
Beat (Hosted by Andrea Bracciali) is completing his PhD research on techniques in distributed computing, data visualization and bio-informatics tools. He developed his project jointly with the Institute of complex systems, University of Western Switzerland, and Universitaat Wurzburg, under the supervision of Prof. Pierre Kuonen, University of Applied Sciences of Western Switzerland. | How computer science can help to make DNA analysis more accessible . In recent years the world of genetics experienced a revolution through the rise of the so called Next Generation Sequencing (NGS) technologies. They allow to sequence DNA quickly and cheaply to create digital files that can be analysed. Through NGS the amount of data created exploded, requiring new approaches to handle this amount of data, both from a computational graphical user interface perspective. For the specific use-case of NGS in diagnostics we explored how to approach those problems by creating a user-friendly graphical NGS data analysis tool. We present the different types of data analysis that are possible to do with the pipeline as well as how we lowered the computing power required to do some of those analysis. We also present the integration of distributed computing into the pipeline to enable data analysis for people with a restricted access to a good computing infrastructure. In the end we look at upcoming research of how we intend to enable researchers to share data and computing power securely across multiple laboratories. | Friday 10 Feb Room: 4B96 |
Dr Iain J. Gallagher (Hosted by John R. Woodward) completed a PhD in immunology at the University of Edinburgh (2006) before moving fields to work in transcriptomics and human physiology at Heriot-Watt University (2007). From there he moved to a post doctoral position in the Medical School of the University of Edinburgh (2009) where he continued research into human muscle pathology as well as developmental biology, metabolic disease and cancer. Iain briefly returned to immunology with a post doctoral position at the Roslin Institute (2011) before taking up his current post at the University of Stirling (2012). Iain uses transcriptomic technology to examine the pathology of muscle wasting and metabolic disease. He has recently published the first multi-tissue classifier of healthy ageing providing a tool for e.g. enrichment of clinical trials. Iain is a keen user of python and R as analysis tools in his research. He has a developing interest in Bayesian statistical approaches. | The Bayesian approach to statistics, . Most scientists are taught frequentist based null hypothesis significance testing (NHST) as the route to inference when faced with noisy, real world data. Developed by 'student' (WS Gosset), Egon Pearson, Jerzy Neyman and most notably Ronald Fisher NHST has stood as the statistical paradigm for around 80 years. Despite efforts by the originators of the approach the devotion to a p-value of <0.05 as indicating some substantive finding is pervasive. Pre-dating NHST by some 100 years the approach of Thomas Bayes (published posthumously and developed by Laplace) differs both philosophically and mechanically from NHST. Proponents suggest that whilst the Bayesian approach is more mathematically challenging the inference is clearer. Recent developments in computing power and a recognition that NHST contributes to irreproducibility in scientific research has led to an increase in interest in the Bayesian approach. In this talk I will introduce the Bayesian approach to statistics, define some key terms and illustrate the approach by example. | Friday 17 Feb Room: 4B96 |
Julie R. Williamson (Hosted by Carron Shankland): I am a lecturer in HCI at the University of Glasgow. I am part of the Glasgow Interactive Systems Group (GIST), leading the Public and Performative Interaction theme within GIST. My research focuses on how people use technology in public spaces and how interactive technologies can be designed given the performative aspects of using technology in public. My current research looks at playful interfaces for public spaces the use embedded interaction, large format displays, and whole body input. | Engagement with Spherical Displays in Public Spaces . Over the past two years, I have completed a series of playful and performative deployments in public spaces using spherical displays. These displays have unique affordances that make them interesting objects for engagement, allowing groups of users to gather around the display and interact together. During this talk, I will discuss what I mean by engagement in this context, what makes spherical displays special in public spaces, and the implications and limitations of completing this intervention based research in public spaces. | Friday 24th Feb Room: 4B96 |
Don Giovanni Battista Spinelli Barrile di Marianella (Hosted by John Woodward) is a Big Data Engineer and work-in-progress Data Scientist. Employed by Compass GmbH in Germany, he is also attending the University of Stirling where is completing the Master of Science in Big Data. Passionate Pythonista, he has experience with several databases such as MongoDB, Neo4J and different flavors of SQL. Lately he has been diving into Machine Learning with the specific interest of performing sentiment analysis on big data sets. He is currently in charge of the backend development of the Airbnb Monitor: a project focused on studying the mid to long-term effects of the popular peer-to-peer app on the housing and tourism market. He also works on Smena, a project concerning network analysis for data retrieved through social media. | Scrapy: an insight on Crawling Data from the Web with Python . Data Scientists' job rely on having access to (large) datasets. However, the data is not always directly available to be processed and analysed; it often needs to be retrieved initially from external sources. Such sources may be an API (application protocol interface) or even web pages from the Internet. Web Scraping or Web Harvesting is the technique of extracting large amounts of information directly from the source code standing behind web pages. During this seminar we will have a quick overview of the Python tools available to perform Web Scraping. A fter this, our focus will shift to Scrapy, an elaborated, free, open-source, web crawling framework written in Python. After explaining the framework's overall structure and its main components, we will see Scrapy in action during an interactive session; this will be done by coding some scripts that are able to recursively collect information from thousands of web pages. During the last stage of the seminar, we will learn how to select, retrieve and store specific information from web pages. | Friday 3rd Mar Room: 4B96 |
title . abstract | Friday 10th Mar Room: 4B96 |
Dr Simon Rogers (Hosted by John Woodward) is a senior lecturer in the School of Computing Science at the University of Glasgow. He has an undergraduate degree in electrical and electronic engineering and a PhD in engineering maths. Since starting his PhD he has worked on the development of machine learning models for problems in biology, particularly those involving large, complex datasets. He is currently focused on the development of models for metabolomics (the study of the small molecule content of a biological sample) where the combination of large volumes of data and poor scientific models make machine learning a promising approach. This work has been published in many top journals (PNAS, Bioinformatics, etc). He is also the co-author (with Prof. Mark Girolami) of an introductory textbook on Machine Learning: A First Course in Machine Learning. | Mining the fragmentome with topic models . Abstract: Metabolites are small molecules central to the operation of living organisms. Measurement of metabolites is desirable across much of the medical and life sciences. However, metabolite identification from mass spectrometry data (a key step in measurement) is very challenging. One common method is to fragment measured molecules and compare the resulting fragments with reference databases. Unfortunately, reference databases are very small and aren&lsquot getting much bigger anytime soon. In this talk I&lsquoll describe an alternative approach based on the use of text mining algorithms (topic models) to analyse large collections of fragment spectra. I&lsquoll describe the topic modelling framework that we have developed as well as some of the challenges involved in working with this data and show some results that demonstrate the power of this approach. | Friday 17th Mar Room: 2V2 |
Dr Livio Pompianu (Hosted by Andrea Bracciali) is a PhD student at the University of Cagliari (Italy). He is completing his PhD research on programming languages for distributed systems, focusing on blockchains and smart contracts. | An analysis of Bitcoin OP_RETURN metadata . The Bitcoin protocol allows to save arbitrary data on the blockchain through a special instruction of the scripting language, called OP_RETURN. A growing number of protocols exploit this feature to extend the range of applications of the Bitcoin blockchain beyond transfer of currency. A point of debate in the Bitcoin community is whether loading data through OP_RETURN can negatively a?ect the performance of the Bitcoin network with respect to its primary goal. I will present our empirical study of the usage of OP_RETURN over the years. We identified several protocols based on OP_RETURN, which we classified by their application domain. We measured the evolution in time of the usage of each protocol, the distribution of OP_RETURN transactions by application domain, and their space consumption. | Friday 24th Mar Room: 4B96 |
Guevara Noubir (Hosted by Marwan Fayed) holds a PhD in Computer Science from the Swiss Federal Institute of Technology in Lausanne. His research covers both theoretical and practical aspects of privacy, security, and robustness in networked systems. Prior to joining Northeastern University, he was a senior researcher at CSEM SA (1997-2000) where he led the design and development of the data protocol-stack of the third generation Universal Mobile Telecommunication System (UMTS) and its world first 3G prototype. His research led to a wide range of mechanisms and algorithms for scalable, secure, private and robust wireless and mobile communications. He led the winning team of the 2013 DARPA Spectrum Cooperative Challenge. He is a recipient of the National Science Foundation CAREER Award (2005), the ACM Conference on Security and Privacy in Wireless and Mobile Networks (WiSec) best paper award in 2011 and runner-up best paper in 2013. Dr. Noubir held visiting research positions at Eurecom, MIT, and UNL. Dr. Noubir served as program co-chair of several conferences in his areas of expertise such as the ACM Conference on Security and Privacy in Wireless and Mobile Networks, IEEE Conference on Communications and Network Security, and IEEE WoWMoM. He serves on the editorial board of the ACM Transaction on Privacy and Security, and IEEE Transaction on Mobile Computing. | Cross-Layer Attacks in Emerging Networks . The last decade has seen the rise of several new networking technologies, from mobile and wireless to overlay anonymous communication networks such as Tor. In this talk, I will argue that such networks are vulnerable to a variety of cross-layer attacks on their intrinsic features. For instance, an adversary can infer users location using malicious apps without requiring permissions, or by exploiting the physical layer characteristics. I will also provide evidence that the Tor anonymity network is also subject to active attacks, and present a framework that identifies malicious relays. I will then discuss the result of the use of the framework, revealing over 100 malicious relays. | Friday 31st Mar Room: 4B96 |
Foteini Katsarou (Hosted by John Woodward) is a PhD student, towards the end of her studies, at University of Glasgow in the school of Computing Science under the supervision of Prof. Peter Triantafillou and Dr Nikos Ntarmos. She is currently working on graph databases and ways to expedite the subgraph querying problem. Foteini competed her undergraduate and master studies in University of Thessaly, in the Department of Electrical and Computer Engineering. | Subgraph Querying with Parallel Use of Query Rewritings and Alternative Algorithms . Graphs are ideal for representing complex entities and their relationships/interactions. Subgraph querying is essential for graph analytics applications. Thus, it is often required to retrieve those graphs from a given dataset that contain a specific query graph (decision problem) and/or find all its occurrences (matching problem). To address this, many methods have been proposed. Our work is not an ordinary comparison among the various methods. We complement existing work by appropriately exploiting the "hidden-gems" of previous work and by combining them in a novel framework to achieve dramatic performance gains. Firstly, we recognize the existence of straggler-queries. With the existing methods reporting only on workload execution times, this can lead to erroneous conclusions on their performance. Secondly, we study the dramatic effect of isomorphic queries on the query execution times. Thirdly, we introduce a set of query rewritings to achieve large performance gains. Fourthly, we show that straggler queries are algorithm specific, i.e. a straggler query on one algorithm can be a typical query on the other. All these discoveries/observations naturally lead to the derivation of a novel framework, coined ?-framework, for subgraph query processing that employs existing methods in parallel to achieve large performance gains. Specifically, the central idea is the parallel initiation of query rewritings and/or alternate algorithms. The framework is shown to be highly beneficial across algorithms and datasets. | Friday 5th May Room: 4B96 |
Dr.Juan Ye (Hosted by Sandy Brownlee) Juan Ye is a lecturer in School of Computer Science at the University of St Andrews, UK. Her research interests resides in designing and developing knowledge-and learning-hybrid techniques in human activity recognition, with speciality in ontologies, context modelling and reasoning, and uncertainty reasoning techniques. She received her Ph.D. degree from University College Dublin in 2009. She publishes in top venues in pervasive computing, including ACM transactions on Interactive Intelligent Systems, Pervasive and Mobile Computing, KER, IEEE Pervasive, Percom, IEEE SMC, IUI, Mobiquitous. | New challenge and trend in sensor-based human activity recognition . Recognising high-level human activities from low- level sensor data is a crucial driver for pervasive systems that wish to provide seamless and distraction-free support for users engaged in normal activities. Research in this area has grown alongside advances in sensing and communications, and experiments have yielded sensor traces coupled with ground truth annotations about the underlying environmental conditions and user actions. Traditional machine learning has had some success in recognising human activities. However, the potential of deploying pervasive systems in real-world environments has introduced a lot of new challenges, such as dealing with sensor failure, lacking of well-annotated ground truth, dealing with multiple users and recognising their interleaved activities, and detecting and learning unknown activities. This keynote will discuss each of these challenges and introduce our work towards addressing them. | Friday 12th May Room: 4B96 |
Dr Pablo Casaseca (Hosted by Sandy Brownlee) | title . abstract | Friday 19th May Room: 4B96 |
Matteo Cavaliere (Hosted by Andrea Bracciali) | Rise and Fall of Cooperation in Structured Populations: A Test with Synthetic Biology . Social, biological and economic networks grow and decline with recurrent fragmentation and re-formation, often explained in terms of external perturbations. I will present a model of dynamical networks and evolutionary game theory that explains these phenomena as consequence of imitation and endogenous conflicts between "cooperators" and "cheaters": Cooperators promote well-connected prosperous (but fragile) networks and cheaters cause the network to fragment and lose its prosperity. Once the network is fragmented it can be reconstructed by a new invasion of cooperators, leading to recurrent cycles of formation and fragmentation observed, for instance, in bacterial communities and socio-economic networks. I will discuss how this type of models can be helpful to understand the role of cellular cooperation in structured populations and the possibilities to use synthetic biology to experimentally validate these ideas. |
| Previous Seminar Series |
|---|
| 2016: Spring Autumn |
| 2015: Spring Autumn |
| 2014: Spring Autumn |
| 2013: Spring Autumn |
| 2012: Spring Autumn |
| 2011: Spring Autumn |
| 2010: Spring Autumn |
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