Seminar in Data Science WS 2026/27

Course DAT.B62UB Uni-Graz online

Objective:

Students will autonomously investigate topics of their own choosing to extend their knowledge beyond the subjects discussed in the data science studies, deepen their understanding of data science related techniques and learn important skills in literature research, self-organization and the presentation of their findings.

Process:

Subject Proposal:
The students are expected to propose topics of their own choosing. These topics should not be part of the regular data science curriculum but extend to related topics; e.g., not simply "Neural Networks" but more specialized, such as "Issues of Vanishing Gradients for Recurrent Neural Networks"... The investigations should focus on a theoretical aspects of techniques, e.g., what are the principal motivations or ideas for approaches? why do things work? why/when do they not work? etc.
The students must submit a one or two-page proposal of the subject they want to work on by Thursday, October 15 to Moodle. This proposal needs to include:

• background on the subject, motivating questions, etc.
• which are the aspects to be investigated (e.g., why do recurrent networks fail for certain tasks).
• at least two references (scientific articles/books) to the subject. (The students are not expected to have fully read/investigated the mentioned references at the time of the proposal; but they should offer a good starting point to investigate the topic, and other references of interest might naturally come up during the students' research.)

Non-exhaustive list of topics: Data analysis, mathematical models, computer vision, computer science, statistical methods, big data, machine learning, optimization algorithms, inverse problems, etc. For students that have already finished, or are in the process of finishing the Project in Data Science, the same overall topic as in the project can be proposed, however, the concrete work done needs to be sufficiently different (e.g., in the project one used certain algorithms, while in the seminar the capabilities/limitations of those algorithms are investigated on a deeper level).
Proposal defence: On October 16th the students should present their project proposals in a mini presentation (3 minutes, only one slide).

Written Summary:
The students are expected to summarize their findings in a written document of roughly 10-20 pages. This text should be self-contained, i.e., understandable for readers without being an expert in the specific topic, and scientific statements should be underpinned with references to the scientific literature. To ensure the students' progress, there will be two meetings with Richard Huber where the current state of the work will be discussed, and suggestions for improvement are made. The students are not expected to have a perfect written document at that point, but the further the work has progressed the more concrete suggestions for improvement and further guidance can be given.
The students need to submit their final report no later than 31.01.2027 via Moodle.

Presentation:
In January, the students are expected to present their findings to the other participants of the seminar in roughly 30 minute presentations. These presentations are not only meant to showcase what the student has learned, but simultaneously be educational to the listeners, so that students gain an overview of all topics. After each presentation, the presenter is expected to answer the audience's questions on the concrete findings, but also on the research topic more broadly. The sessions for presentation in January are already set, but when which student will present their work will be decided at a later point.

In-Person seminar classes:

The seminar classes will NOT take place every week, but only upon agreement (see Relevant dates below). The meetings that are held in person will take place on Fridays from 08:30 until 10:00 in the seminar room SR 11.34 (0011030012). The presence at the in-person meeting is compulsory and absence will only be excused for cause. The students are of course welcome to contact me via richard.huber@uni-graz.at with questions or ask for in-person meetings if assistance is required.

Progress discussions:

There will be two progress-discussion sessions (see dates below) in which students will have one-on-one meetings with Richard Huber to describe what progress was made, in which direction things are developing, what their results are etc. These meetings are not graded and are meant as an opportunity for the students to gain additional feedback and for Richard Huber to monitor (but not grade) the progress made.

Grading:

As mentioned above, the seminar has three principally graded activities: the subject proposal and defence (1/6), the written report (1/2) and the presentation (1/3). The final grade is gained as a weighted average of these sub-grades with one-sixth, one-half and one-third contributions, respectively. Additionally, each part must be positive by itself, i.e., the final report must be satisfactory even if performance in the presentation was excellent. The progress discussions are not graded, but participation is compulsory.

Relevant dates:

Date	Subject	Remarks
02.10.2026	Discussion of the seminar format and grading
16.10.2026	Presentation of topic proposals
13.11.2026	First progress discussion	Will not take place in the seminar room, but in Leechgasse 34. The precise times can be agreed with Richard Huber upon request.
11.12.2026	Second progress discussion	Will not take place in the seminar room, but in Leechgasse 34. The precise times can be agreed with Richard Huber upon request.
08.01.2027	Presentation of results
15.01.2027	Presentation of results
22.01.2027	Presentation of results
29.01.2027	Presentation of results