Details Colloquium Fall 2015

24.09.2015 - Mining Software Data for Mobile Apps: Achievements, Challenges, and Research Direction

Speaker: Prof. Dr. Massimiliano Di Penta
Host: Prof. Dr. Harald Gall  

Abstract

In recent years we are assisting to a wide diffusion of mobile applications, with the apps' economy becoming a great success and an extremely lucrative source of revenues for developers. From a perspective of researchers interested to provide support to mobile app developers, it would be extremely interesting and challenging to understand how to tailor mining software repository techniques to this complex yet promising and emerging area, and what kinds of recommenders could be relevant for mobile app developers.On the one hand, this area provides a lot of promising opportunities, among others the availability of app reviews and ratings, that can be extremely useful to build recommenders for project managers and analysts. Also, mobile apps pose very peculiar software analysis challenges, for example related to validating them against multiple devices, or identifying opportunities to reduce energy consumption. On the other hand, many of these applications are developed by very small teams, if not by a single person, without often adopting a fully-fledged software configuration management, hence limiting the availability of resources to be used when mining their evolution history. Also, recent studies suggested that the analysis of apps' source code or byte code require particular attention, because of the massive usage apps are using of particular resources, such as advertisement brokers or providers. This talk will overview recent achievements in the area of mining mobile apps, and will highlight achievements and challenges in this research area.

Bio

Prof. Dr. Massimiliano Di Penta is associate professor at the University of Sannio, Italy since December 2011. Before that, he was assistant professor in the same University since December 2004. His research interests include software maintenance and evolution, mining software repositories, empirical software engineering, search-based software engineering, and service-centric software engineering. He is author of over 200 papers appeared in international journals, conferences and workshops. He serves and has served in the organizing and program committees of over 100 conferences such as ICSE, FSE, ASE, ICSM, ICPC, GECCO, MSR WCRE, and others. Working Conference on Mining Software Repositories (MSR 2015). He has been general chair various events, including the 12th Working Conference on Mining Software Repositories (MSR 2015), the 10th IEEE Working Conference on Source Code Analysis and Manipulation (SCAM 2010), the 2nd International Symposium on Search-Based Software Engineering (SSBSE 2010), and the 15th Working Conference on Reverse Engineering (WCRE 2008). Also, he has been program chair of events such as the 28th IEEE International Conference on Software Maintenance (ICSM 2012), the 21st IEEE International Conference on Program Comprehension (ICPC 2013), the 9th and 10th Working Conference on Mining Software Repository (MSR 2013 and 2012), the 13th and 14th Working Conference on Reverse Engineering (WCRE 2006 and 2007), the 1st International Symposium on Search-Based Software Engineering (SSBSE 2009), and other workshops. He is currently member of the steering committee of ICSME, MSR, SSBSE, and PROMISE. Previously, he has been steering committee member of other conferences, including ICPC, SCAM, and WCRE. He is in the editorial board of IEEE Transactions on Software Engineering, the Empirical Software Engineering Journal edited by Springer, and of the Journal of Software: Evolution and Processes edited by Wiley.

08.10.2015 - Ontology Lexicalization: A Core Task for the Semantic Web

Speaker: Prof. Dr. Philipp Cimiano
Host: Prof. Dr. Martin Volk

Abstract

In this talk we argue that access to the Semantic Web will be mediated by natural language for most casual users of the World Wide Web. Thus, knowledge about how vocabulary elements of ontologies used in the Semantic Web are verbalized in different languages is key to support language-mediated access to the Semantic Web. We present the lexicon model for ontologies (lemon) that allows to represent information about how ontology elements are verbalized in different natural languages. Further, we introduce a new task we call ontology lexicalization that consists in automatically inducing lexicalization candidates for Semantic Web ontologies, thus supporting the semi-automatic development of lexica in lemon format. We present a new approach to ontology lexicalization called MATOLL and describe experimental results on the ontology lexicalization task with respect to the DBpedia ontology for three natural languages: English, Spanish and German. We describe a resource called DBLexiPedia that releases lexicalizations for these three languages for the DBpedia ontology as Linked Data. Finally, we discuss potential applications for lemon lexica in the context of Question Answering for the Semantic Web.

Bio

Prof. Dr. Philipp Cimiano is full professor of computer science at Bielefeld University. He is head of the newly founded Semantic Computing group and is affiliated to the Cluster of Excellence on Cognitive Interaction Technology (CITEC). Philipp received his doctoral degree in Applied Computer Science from the University of Karlsruhe (now KIT). His PhD research was on the topic of learning ontologies from text and led to the publication of a book at Springer Verlag with the title “Ontology Learning from Text: Algorithms, Evaluation and Applications”. He has published over 120 research papers in the areas of Natural Language Processing, Ontology Learning, Ontologies, Knowledge Acquisition and Semantic Web. Philipp was nominated one of “AI’s 10 to Watch” by the IEEE Intelligent Systems Magazine in 2008, an award given to the top 10 young researchers in the field of artificial intelligence worldwide. His most recent book has the title "Ontology-based Interpretation of Natural Language" and has been published by Morgan & Claypool in the series "Synthesis Lectures on Human Language Technologies".

19.11.2015 - Monitoring in Large Systems

Speaker: Metin Feridun, Ph.D.
Host: Prof. Dr. Burkhard Stiller

Abstract

Monitoring is a core function of systems management. The status of systems elements and associated performance metrics are collected, filtered, aggregated and analyzed to determine the health of the system and diagnose problems. Although the concepts of monitoring are well known and understood, a number of issues always provide challenges in systems management design, such as the integration of data from heterogeneous systems and scalability. These two topics are particularly important today as they need to be addressed for large scale systems such as cloud computing and distributed file systems. This talk will summarize the key concepts of monitoring in systems management and provide an overview of the challenges faced in monitoring large scale-systems. It will then survey some of the solutions that are available today and provide an in-depth study of the design and implementation of a monitoring solution for such a system. The talk will conclude with a discussion on approaches to health monitoring in large systems.

Bio

Metin Feridun, Ph.D., is a researcher at the IBM Research- Zurich based in Rueschlikon since 1990. His focus of research is distributed systems management, specially monitoring, event management, and integration. He holds a PhD from Cornell University (1984) and an MBA from University of Strathclyde (2006).

26.11.2015 - Causal Analysis for Association Discovery

Speaker: Prof. Dr. Jiuyong Li
Host: Prof. Dr. Daning Hu

Abstract

Association analysis is an important technology in data mining, and has been widely used in many application areas. However, associations in data can be spurious and they do not indicate causal-effect relationships that are ultimate goals for many scientific explorations and social studies. While the techniques for association discovery become mature, the problem for identifying non-spurious associations becomes prominent. In this talk, I will discuss two methods for testing causal effect relationships rooted in statistical analysis and the use of them for filtering non-causal associations in association discovery. The first method uses the Mantel and Haenszel test for partial association, and the second method uses a cohort study approach. I will also discuss the implementation issues in large data sets for combining relationship exploration and evaluation.

Bio

Prof. Dr. Jiuyong Li is a professor and an associate Head of School at the School of Information Technology and Mathematical Sciences of University of South Australia. He leads the Data Analytics Group in the School. His main research interests are in data mining, bioinformatics, and data privacy. He has led multiple Australian Research Council Discovery projects. He has published more than 100 papers, mostly in leading journals and conferences in the areas. His software tools have been used in several real world projects. He has been chair (and PC chair) of Australasian Data Mining Conference and Australasian Joint Conference on Artificial Intelligence. He has received senior visiting fellowships from Nokia Foundation, the Australian Academy of Science, and Japan Society of Promotion of Science.

03.12.2015 - Linking Everything

Speaker: Prof. Dr. Manfred Hauswirth
Host: Prof. Dr. Harald Gall

Abstract

We produce humongous amounts of information - technical infrastructures (smart grid, smart cities), the Internet of Things (sensors), mobile phones, the Social Web (Twitter, social networks, blogs), the media, and many more - and these amounts are growing exponentially. Linked Data gives us the technical means to network all this information and enables us to develop new forms of analytics on networked data from many sources instead of traditional "monolithic" data analytics. But this network of information is "in-discrete" as the data is produced continuously and at potentially high speeds with varying loads and demands on the producer and the consumer sides. This calls for new data/knowledge management approaches and as a result, the Linked Data world is moving from a simplifying discrete model to a more realistic continuous view. This development impacts on and changes research problems in all areas and for all layers and requires well-orchestrated research efforts in and across research communities to support "streaming" as an integrated paradigm. In this talk, I will present a comprehensive stack of Linked Stream management approaches for all layers - from the Internet of Things to backend information systems, and will discuss the impact of streams on big data, analytics, and privacy.

Bio

Prof. Dr. Manfred Hauswirth is the managing director of Fraunhofer FOKUS since October 2014 and a full professor for Open Distributed Systems at the Technical University of Berlin since September 2014. Before that he was the Vice-Director of the Digital Enterprise Research Institute (DERI), Galway, Ireland and a professor at the National University of Ireland, Galway (NUIG) for 8 years and a stream leader in INSIGHT, the national Irish data analytics centre. He still holds a position as research professor at NUIG. Prior research positions include Ecole Polytechnique Federale de Lausanne (EPFL, 2002-2006) and TU Vienna (1999-2002). He holds an M.Sc. (1994) and a Dr. degree (1999) in computer science from the Technical University of Vienna. His research interests are on Internet-of-Everything, domain-specific Big Data and analytics, linked data streams, semantic sensor networks, sensor networks middleware, large-scale semantics-enabled distributed information systems and applications, peer-to-peer systems, service-oriented architectures and distributed systems security. He has published over 160 papers in these domains, he has co-authored a book on distributed software architectures and several book chapters on data management and semantics. He has a long track record of working wiht industry specifically in the IoT, business process, telecoms, networking and data integration domains. He has served in over 190 program committees of international scientific conferences and was program co-chair of the Seventh IEEE International Conference on Peer-to-Peer Computing (IEEE P2P) in 2007, general chair of the Fifth European Semantic Web Conference (ESWC) in 2008, program co-chair of the 12th International Conference on Web Information System Engineering (WISE) in 2011, and program co-chair of the 10th International Conference on Ontologies, DataBases, and Applications of Semantics (ODBASE) in 2011. He is a member of IEEE and ACM and was on the board of WISEN, the Irish Wireless Sensors Enterprise Led Network, the scientific board of the Corporate Semantic Web research center at FU Berlin, and the Scientific Advisory Board of the Center for Sensor Web Technologies (CLARITY) in Dublin, Ireland.

10.12.2015 - Quadratic-Price Combinatorial Auctions

Speaker: Sèbastien Lahaie Ph.D.
Host: Prof. Dr. Sven Seuken

Abstract

I present an iterative combinatorial auction that offers modularity in the choice of price structure, drawing on ideas from convex optimization and machine learning. The auction can incorporate the standard item- or bundle-pricing schemes, but also any degree of polynomial prices in between, maintaining a sparse representation of the prices throughout. The design is motivated by a theoretical analysis which indicates that low-dimensional (e.g., item) prices lead to dramatically faster convergence than high-dimensional (e.g., bundle) prices, a phenomenon already observed in practice. An empirical evaluation reveals that quadratic price auctions are sufficient to clear the market across standard benchmarks, and remain competitive against state of the art ascending-price auctions in terms of efficiency and revenue. During the talk, I will highlight the similarities and differences between the problem of market clearing and classical learning problems like classification and regression.

Bio

Sébastien Lahaie, Ph.D., is a senior researcher and founding member at Microsoft Research in New York City. He received his Ph.D. in Computer Science from Harvard in 2007 and was previously a senior research scientist at Yahoo. His research focuses on computational aspects of market design, with applications to sponsored search and display advertising. He is interested in designing market algorithms that scale well and properly anticipate user behavior. Other interests include preference modeling and elicitation, combinatorial auctions, and prediction markets. He serves as a co-editor for Economic Inquiry, was previously a program chair for AMMA, and has served on the senior program committee of conferences such as EC, AAAI, and WWW.