Current Topics in Information-theoretic Data Mining

data mining by data compression

Latest News

  • Submission format: The report, as well as your presentation file in *.pptx format, must be emailed to your supervisor at or before 09:00 on the 15th May (the first day of the seminar).
  • Detailed seminar agenda now available (see below).
  • The templates for the report and presentation are available for download below.
  • To all students that have not yet contacted their supervisor: Do so as soon as possible!
  • To all students that do not yet have a copy of their research paper: email your supervisor and request a copy.
  • Slides from kick-off meeting (24th January 2014) are here.

Seminar Dates / Location / Presentation Agenda

Day 1 of 2: Thursday May 15, Room MI 01.07.023

09:00 Benjamin G. (Supervisor: Annika)
09:45 Tim B. (Supervisor: Annika) - short break to follow
10:30 Jan G. (Supervisor: Annika)
11:15 Veronika K. (Supervisor: Annika)
12:00 Lunch
13:00 Kathrin R. (Supervisor: Annika)
13:45 Martin K. (Supervisor: Nina) - short break to follow
14:30 Markus L. (Supervisor: Nina)
15:15 Gabriel W. (Supervisor: Nina)
16:00 Finish Day 1

Day 2 of 2: Friday May 16, Room MI 02.13.010

09:00 Laura A. (Supervisor: Nina)
09:45 Christian-Friedrich L. (Supervisor: Nina) - short break to follow
10:30 Simon G. (Supervisor: Sam)
11:15 Philipp D. (Supervisor: Sam)
12:00 Lunch
13:00 Christian B. (Supervisor: Sam)
13:45 Benedikt B. (Supervisor: Sam) - short break to follow
14:30 Zeeshan S. (Supervisor: Sam)
15:15 Seminar wrap-up and conclusion


  • Presentation Template: Participants are required to use this PowerPoint template for their presentation. Note that the existing slides in the template are simply to illustrate various formatting possibilities (i.e. they should NOT be interpreted as any kind of outline-structure for your presentation).
  • Report Template: Participants are required to use the IEEE Manuscript Template for Conference Proceedings. Windows download (ZIP), Unix download (TAR.GZ).


Recent information-theoretic approaches relate data mining to data compression: Data mining techniques detect patterns or regularities in data. The more effective we can compress the data with the knowledge of the detected patterns, the more novel information we have learned from our data. Many research papers demonstrate the potential of this very general idea: Information-theoretic techniques support parameter-free data mining, i.e. the user does not need to specify any input parameters which are difficult to estimate. Information-theoretic methods support a wide range of data mining tasks including clustering, outlier detection, and association rule mining on various types of data, e.g. numerical feature vectors, categorical data and graphs. Every student selects a recent research paper which he/she presents in the seminar. This seminar is ideal for students interested in doing a Bachelor- or Master's thesis on related topics within my group iKDD or for those with a general interest who want to learn more.


Basic knowledge in data mining and/or machine learning is helpful but not mandatory. The requirements for a passing grade in the seminar are:

  • Giving a 30 minute presentation (including discussion) of your topic.
  • Writing a seminar paper. The length should be approximately 3-4 (Bachelors seminar) to 5-6 pages (Masters seminar) in IEEE layout (see 'Resources' above). The report, as well as your presentation file in *.pptx format, must be emailed to your supervisor at or before 09:00 on the 15th May (the first day of the seminar).
  • Meeting at least once with your advisor before giving the presentation.
  • Participation during the introductory session and all block seminar days.

Organization and Important Dates

  • Block seminar: 15-16 May 2014. Electronic versions of the presentation and report are due at 09:00 on the 15th May as stated above.
  • Kick-off meeting to inform about and register for the seminar: Friday, January 24th, 10.00 am, room MI 03.11.018