: Open Theses
: uni gr
Best Poster AwardMihail Minev received a Best Poster Award at the PROMISE Winter School 2013 titled Bridging between Information Retrieval and Databases.
PhD DefenseJayanta Poray: INACS - An Incremental and Adaptive System for Conversational Streams.
Thursday, December 13, 14h00 - Salle de Conseils.
Master DefenseZhaochun Ren: Topic Integration and Summarisation in Web Forums.
Tuesday, November 13, 2012.
Master Defense (Co-Reviewer)Diana Marosin:
Friday, September 14, 2012.
Master DefenseSergio Sousa: Intention Detection For BCI Controlled Robots.
Thursday, July 19, 14h00 - Salle de Conseils
Research Meeting & PhD Day, Miami UniversityWe participated the research day, organised by the Dica Lab, U Luxembourg.
TalkTrusted ICT for Finance Workshop : April 23, 2012, Luxembourg School of Finance - 2B, rue Albert Borschette, Groundfloor
16.30 - 17.00 Data Mining in Finance. Prof. Christoph Schommer, Mihail Minev, CSD, University of Luxembourg
Pictures from ICAART 2012 Conference (copyright: INSTICC) of the MINE group members
Computational Finance Working Group Meeting
At the UL Christmas Party - December 13, 2011
from left to right: Prof. Kelsen, D. Shirnin, Prof. Schommer, F. Schmitz, M. Minev
VisitWho: Prof. Dr. Stephan Busemann, German Research Centre for Artificial Intelligence, Saarbruecken.
When: Wednesday, November 16, 2011.
VisitWho: Dr. Ulrich Schaefer, German Research Centre for Artificial Intelligence, Saarbruecken.
When: Wednesday, September 28, 2011 at 10h00 in Room E012
Internal Research Meetings 2011 (MINE Research Day)When: September 14, 2011, from 10h00 until 15h00.
Internal Research Meetings 2011 (MINE Research Day)When: June 29, 2011, from 11h00 until 15h00.
Where: Room E.012
TalksMay 12, 2011, 08h15 - 10h30: Building C - Room C11: Applications of Data Mining
TalkApril 27, 2011, 14h00 : Building F - Room 213 (SnT)
Fabien Bouleau, SES Engineering: Talk about "Anomaly Detection and Spacecrafts"
Spacecrafts provide a large set of on-board information about the health of the different components and subsystems, that are constantly monitored by Satellite Controllers via a ground control system software (GCS). This software retrieves the data usually coming down as frames of bytes that are processed in real-time using different mechanisms to be displayed in a human readable way under the form of parameters. All the frames are in parallel being collected in different a database (that can be of different type, depending on the GCS) to make it available for offline post-processing. The GCS is able to inform the Satellite Controller when parameters values are deriving to abnormal values, base on thresholds (upper and lower out-of-limits definitions). It is nevertheless not able to correlate the data and analyze more complex situations that can lead to anomalies. Mining the historical spacecraft data could help finding correlations between the different type of components (such as power, thermal, solar illumination and orientation, as well as power and thermal on the other hand).
Master Thesis DefenseApril 27, 2011, 11h00
Francois Beckius: A Method for Closed-Loop Brain-Computer Interface Training Using a 14-channel EEG.
Brain-computer interfaces (BCI) allow a user to control an external device merely by thinking about it. They achieve this goal by observing the electrical potential generated by neurons firing within the user's brain. In this master thesis, the amplitude of this brain activity is measured through non-invasive means, using a fourteen-channel electroencephalograph (EEG). The recorded data is fed to the BCI's machine-learning software, which translates the input into useful commands. Thus, the BCI serves as a new input device that can be used to render computers more accessible to disabled patients or can potentially control prostheses directly with brain activity.
One of the great barriers preventing the wide-spread use of this technology is the training time required for users to become proficient at using a BCI. This can require several hours of training a day for multiple weeks. The purpose of this master thesis is reducing the training time and effort required by improving the software used to interpret the brain signals.
The Master thesis currently focuses on making the user able to move a cursor into four directions with his mind. In this approach, the initial phase of the training takes less than a minute, during which data is recorded that already allows the user some imprecise control over the cursor. The remaining training takes place naturally in the background while the user is actually working with the device.
ParticipationRelais pour la vie 2011 - join us (March 12 & 13)!
LCSB - CSC Workshop - January 10 & 11, 2011, Castle Schengen, Luxembourg
Presentation & DiscussionDecember 21, 2010
Dr. Maria Biryukov: Text Mining and Systems Biomedicine
Time and Place: Thursday, December 9, 2010 - 10h00 s.t. - LCSB, Campus Limpertsberg
The permanent growth of the electronically available publications in biomedicine and computational biology makes it difficult to follow the evolution of research and stay updated on the recent developments in the field. That is the reason why text-mining systems become an active area of multidisciplinary research that brings together specialists in biomedicine, bioinformatics and computer scientists with expertise in data and text mining. This is illustrated by Genomics Track of the TREC competitions, sponsored by NIST, and other similar events which test the state-of-the-art text mining and information retrieval technologies against the biological data. The goal of this talk is to provide an overview of the biological text mining and discuss a number of tasks that could be addressed in relation to the research conducted in LCSB and its collaborating institutions.
PhD DefenseSascha Kaufmann: CUBA - Conviviality and User Behavior. DECEMBER 10, 2010. 14h00. Salle de Conseils. Campus Kirchberg.
Abstract: Conviviality is a concept of great depth that plays an important role in any social interaction. A convivial relation between individuals is one that allows the participating individuals to behave and interact with each other following a set of conventions that are shared, commonly agreed upon, or at least understood. This presupposes an implicit or an explicit regulation mechanism based on consensus or social contracts and applies to the behaviours and interactions of participating individuals. With respect to an intelligent web-based system, an applicable social contribution is the give of assistance to other users in situations that are unclear and in guiding him to find the right decision whenever a conflict arises. Such a convivial social biotope deeply depends on both implicit and explicit co-operation and collaboration of natural users inside a community. Here, the individual conviviality may benefit from "The Wisdom of Crowds", which fosters a dynamic understanding of the user's behaviour and a strong influence of an individual's well being to another person(s).
The web-based system CUBA focus on such a behavioural analysis through profiling and demonstrates a convivial stay within a web-based feed system.
Presentation & DiscussionDecember 2, 2010
Dr. Ulrich Schaefer: "Semantic search and visual navigation in scientific publications"
Language Technology Lab, German Research Center for Artificial Intelligence (DFKI), Saarbruecken
Time and Place: Thursday, December 2, 2010 - 11h00 s.t. - Room E012 (Building E, basement, Campus Kirchberg)
Scientists in all disciplines are nowadays faced with a flood of new publications every day. In addition, more and more publications from the past become digitally available and thus even increase the amount of data. Therefore, finding relevant information and avoiding redundancy and duplication of work have become urgent issues to be addressed by the scientific community. The organization and preservation of scientific knowledge in scientific publications, vulgo text documents, thwarts these efforts. From a viewpoint of a computer scientist, scientific papers are just `unstructured information'. We report on recent advances in the project TAKE on the scientist's Workbench, a platform demonstrating the use of natural language processing (NLP) to support scientists or other knowledge workers in fast and better access to digital document content. We focus on sentence-wise NLP analysis of electronically available scientific paper texts. With the generated NLP annotations, we are able to provide two important, novel applications, (1) robust semantic search in paper content and (2) document navigation based on citation classification. In contrast to previous work, we now provide semantic search on the full textual content of scientific papers instead of just abstracts and use a fall-back strategy to maximize NLP robustness on every sentence. We extend the workbench with a novel visualization and navigation facility that aims at helping scientists to quickly adapt to research fields and topics. It is based on NLP-assisted classification of citation sentences to compute a browsable, typed citation graph.
PhD DefenseMaria Biryukov: Methods for Extracting Meta-Information from Bibliographic Databases. NOVEMBER 15, 2010. 15h00. Salle de Conseils. Campus Kirchberg.
Abstract: Due to intensive growth of the electronically available publications in the last few decades, bibliographic databases have become widespread. They cover a large variety of knowledge fields and provide a fast access to the wide variety of data. At the same time they contain a wealth of hidden knowledge that requires steps of extra processing in order to infer it. In this work we focus on extraction of such implicit (or meta) knowledge from the research bibliographic databases by looking at them from sociolinguistic, text mining and bibliometric perspectives. We choose the Digital Library and Bibliographic Database --- DBLP as a testbed for our experiments.
In the framework of the sociolinguistic analysis we build a statistical system for the language identification of personal names. We show also that extension of a purely statistical model with the co-authors network boosts the system's performance. There are several premises motivating our work. For example, it has been shown that the geographical proximity influences research. Moreover, research is constantly evaluated on the national and international basis. To make these and similar investigations less laborious in terms of human effort, ability to automatically assign personal names to the appropriate language seems to be useful.
In the text mining scenario, we perform a number of experiments that focus on topic identification and ranking. While our topic detection approach remains generic and can be used for any kind of textual data, the topic ranking metrics are built upon the information provided by the bibliographic databases. With respect to the topic ranking, our study aims at finding the ways of different topic ordering depending on the question that has to be addressed: in some cases we have to know what are the ``hot" research directions that attract a wide audience, while in some other cases, narrow and probably highly specialized topics have to be put forward. Our results show that one can achieve this goal by varying the type of bibliographic information used to compute the topic rank.
It is popular nowadays to bring technics from bibliometrics and scientometrics into the world of bibliographic databases to investigate the collaboration patterns and explore mechanisms underlying community development. The goal of our bibliometric study is to create a researcher's profile on DBLP and analyze some of the research communities defined by the different conferences, in terms of the publication activity, interdisciplinarity of research, collaboration trends and population stability. We also aim at exploring to what extent these aspects correlate with the conference rank.
Each of the above topics constitutes a method of meta information extraction from bibliographic databases. Moreover they can be used in combination in order to provide better results and in order to obtain a multidimentional view on research and research communities.
Such techniques can also be applied to other similarly structured data sources.
Knowledge TransferMeet us at the Student Fair that takes place at November 11 and 12.
Knowledge TransferWe participate the Luxembourgish Event on "Chercheurs a l'ecole" - which is organised by the FNR.
Lycee Technique Michel Lucius (157, av. Pasteur, Limpertsberg)
Rosy, a girl from the NAO Family, will accompany us.
Scientific ExchangeOctober 19, 2010
Visit of the research groups of Prof. Ziegler and Prof. Max, both from the Dica Lab U Luxembourg, FLSHASE, Campus Walferdange. In this meeting, both teams confronted and discussed boundary-crossing research approaches based on current PhD research work.
To know more about the Dica Laboratory, see here
Presentation & DiscussionOctober 15, 2010
Prof. Dr. Alain Krief, University Namur, Belgium, has visited the MINE Research Group.
Presentation & Discussion10th of September 2010: Internal Research Day with topics about
Talk13th of July 2010: Ronny Heinz presents the concept of his PhD Thesis (Artificial Companions).
Event4th Leonardo Summer School
My team from left to right with Adrien, Charline, Nathalie, Michele, Caroline, and Raphael (and the organiser of the Summer School, Prof. Dr. Massimo Malvetti).
Talk8th of July, 2010: Francois Beckius explains current achievements that have been presented at the (39th Conference on Neural Interfaces) in Long Beach, CA, USA.
Talk30th of June, 2010: Mike de Dood will defend his Bachelor Thesis about "Digital Image Classification of Space Galaxies".
TalkMay 7, 2010: Jayanta gives a talk about Using Explorative-Adaptive Mind-Maps to Recommender Trust at the Luxembourgish-Polish Workshop on Security, Reliability and Trust.
TalkMay 6, 2010: Christoph gives a talk about Data Mining in Security Applications at the Luxembourgish-Polish Workshop on Security, Reliability and Trust.
TalkApril 13, 2010: Mike de Dood will give an intermedidate talk of his Bachelor thesis titled with Digital Image Classification of Space Galaxies (14h30, Room E.012).
Paper about Fraud Detection in Telecommunication DataThe following paper concerns with the mining of telecommunication data: Discovering Fraud Behaviour in Call Detailed Records. It has been published in the Book "Mining your own Business - Telecommunication" (see publication list).
Relais Pour la Vie 2010We participate this event - come on join!
ILILAS News: Nao Computers ArrivedThese are Rosy, Lucy, Jempi, Fonsi and Marvin. See the video of the NAO party: http://robolab.gforge.uni.lu/
Defense - Master ThesisRonny Heinz: Do you know who you are? Profiling Authors on the basis of published paper summaries.
29th of January, 2010: 10h00, Room B02
Nowadays we have access to a huge amount of articles and documents over the internet. This gives rise to different problems: plagiarism, search results containing multiple times the same document, inefficient search using keywords, etc. In this Master's Thesis we try to address some of the problems in the case of the search for scientific publications. We use the author as an important source to find new publications which are of interest to us. Instead of using the publications themselves, we use the abstracts of the scientific publications from a given author to generate a profile for this author. Each of the abstracts is transformed into a graph. The profile is then generated using methods performing different operations on the graphs corresponding to abstracts of the given author. The profile can then we used to find similar authors or to match an ACM category to the author.Our experiments show that we can successfully find similar authors and assign ACM categories. Additionally they show that the profile must not be too broad and not too restricted. We implemented the system in a workbench, where the user can experiment with the different operations and methods to create the profile. Through the use of such a system, we could improve the search for scientific publications by returning not only the publications of an author but also similar authors. Furthermore giving the authors an ACM category, it is possible to easily search for an author in a specific category.
TalkChristoph Schommer: Data Mining and Cognition. Information Day, European Commission. Luxembourg, 14 January 2010.
We understand Data Mining/Knowledge Discovery as a cognitive process, which is a life-cycle of Signals, Patterns Noise, Relevances, Surprise and Disappointment, and Consequences. Cognitive Components are inherent, for example Learning, Interpretation, Creativity, Interaction, Knowledge Use, and Strategy Finding. The talk summarizes this ideas and gives three examples: Adaptive Management of Artificial Thoughts, Conviviality within/between Individuals and Groups, and an Open Information Addressing with Steganography.
TalkFrancois Beckius: The Search for Topic Signatures
I will be presenting a summary of the work I have done on sample data received from the ESPON project in order to extract topic signatures using the WordStat tool. The main points are: approaches taken, problems encountered, results, and possible improvements.
TalkRonny Heinz: Do you know who you are? - Creating author profiles by abstracts of scientific publications.
The Invited Talk concerns the profiling of authors by using the abstracts of scientific publications. For that, the abstract is transformed into a individual graph-based framework, which is achieved by a couple of operations on graphs. Using ACM classification as an ontology backbone, a better understanding of the authors' interests can be achieved.
TalkMaria Biryukov: Mining The Bibliographic Database DBLP
TalkSascha KAUFMANN : e-Conviviality in Web Systems by the Wisdom of Crowds
We motivate the idea of e-conviviality in web-based systems and argue that a convivial social being deeply depends on the implicit and explicit co-operation and collaboration of natural users inside a community. We believe that a (individual) conviviality benefits from the wisdom of crowds, meaning that a continuously and dynamic understanding of the user's behaviour can heavily influence the individual well being. In this respect, we are currently implementing the system CUBA (conviviality and user behaviour analysis), which aims to find novel ways to support users during their web site Invited Visits while discovering their interests. CUBA comes up with certain recommendations and suggestions, which base on a common behaviour of the "Wisdom of Crowds". For example, concepts with respect to time, space, and user-based actions are considered.
TalkJayanta Poray: Towards a healthy Explorative Mind-map framework and beyond
The extraction of patterned information from the raw data source is the key to set-up an artificial cognitive system. In this regard the design of an explorative mind-map for the conversational text streams and ensure its healthy status is a challenging task, involves many machine learning steps. Also the study of human mind theory may influence the overall process. Finally the structured mind-map may used for many real applications. In our naive approach we are formalizing the computational trust with the help of several mind-map elements and their derivatives.
Certification Course - DB2
30th of April, 2009 - UL, Campus Kirchberg
MICS - Newspaper Report about the Master of Science
MINE Research Days, December 16, 2008 - UL, Campus Kirchberg.
MINE Research Days, 9th November 2007, Campus Kirchberg - Room E012
MINE Research Days, 06. July 2007 - University of Applied Sciences, Trier.
From left to right: Michael, Conny, Maria, Sascha, Ejikeme, Christoph, Svetlana, Ralph
March 15, 2007 - March 16, 2007, Room 612, Robert-Mayer-Str. 10-12 (Mathematics Building, Campus Bockenheim). This research day takes place in Frankfurt/Main, Germany (JW Goethe University). It is hosted and organized by the Goethe AG.
Research DayNovember 7-8, 2006, Campus Walferdange.
MINE Research Days, 03 - 04. July 2006 - UL, Campus Kirchberg
MINE Research Days, 09. March 2006 - UL, Campus Limpertsberg.
November 14, 2005 : 14h00 - 19h15 - University of Luxembourg, School of Finance
AI Lecture Series I
Open Doors 2005
Student Party 2005
ILIAS Team 2004