Writing a thesis in the Theory of Machine Learning Group

If you are interested to write a Bachelor or Master thesis in our group, then please do the following:

• Before writing a thesis with me, you should have attended some relevant classes in the field of machine learning. On the Bachelor level, this might be the Introduction to ML, on the master level the Statsticial Machine Learning.
• As the name of our group suggests, much of our research is inspired by theory, but often also requires implemenations and playing with algorithms . We supervise both empirical topics and theory topics --- but you shouldn't be completely scared of maths.
• Our main field of research is on explainable machine learning. Some of our group members also work on language models, so we might have topics in this direction as well.
• Sometimes we have a list of open thesis projects here (same login and password as for all my lectures). But keeping this list up to date turned out challenging. So if there is nothing on this list, or it looks outdated, then I suggest that you look at some of our recent papers and judge whether you find them interesting. If yes, then please get in touch with us as follows:
  Send an email to Ulrike von Luxburg with your relevant background information:
  • Your name
  • Short CV
  • what is your current degree program (Studiengang, bachelor or master), which semester
  • Which relevant courses you took so far in the field of machine learning: which course, when, who was the lecturer, which grade?
  • All your transcripts of records (of your current degree and of all prior degrees / subjects that you have studied)
  • Time line: when would you like to start
  • Which of our current papers you find interesting, or in which area you might consider writing a thesis. For us it is also interesting to know whether you would be rather interested in implementation or in theory
  • We will then let you know whether we have capacities to supervise theses, and if yes, have a meeting in which we make suggestions for a topic.
• If you have an idea for an alternative topic yourself, we can talk about it, but experience shows that often this does not go well. Hence, I am rather reluctant. If you want to try it, I request that you write an outline of about three pages about your topic: what is the question you want to solve, why is it interesting, what are the methods you would like to try, what literature do you know, etc. We then meet to discuss your proposal, and only after this meeting I am going to decide whether I am going to supervise this topic or not.
• Note that the process to discuss a topic and agree on supervision (or not) typically takes a couple of weeks, so do not wait until the last minute.
  
  

• Even though I teach Bachelor lectures on algorithms and theoretical computer science, this is not really my field of research. For this reason I do NOT offer thesis topics in theses areas (you might want to ask Michael Kaufmann or Britta Dorn or Lena Schlipf instead). We only supervise topics in machine learning.
• I do NOT supervise an external thesis in some industry company. Often, industry projects do not have enough scientific content to serve as a thesis, and I have too little control about what is going on. Note also that our research focus is on theory, not on applications.