Heidelberg University

Seminar: Statistics of Generative Models

Wintersemester 2024/2025

For any questions regarding contents and organization, please get in touch via mail.

Links: MUESLI; Moodle (The Moodle page will be made accessible a couple of days prior to the first seminar.)

Description

The seminar will consist of student presentations about current research papers on the statistics of generative models. This is not a seminar on applied generative models, e.g. in image generation, but on utilizing arguments from mathematical statistics for investigating processes in generative modelling. Accordingly, some prior knowledge on mathematical statistics, e.g. Statistik 1, will be required. The seminar will be in a block format with two initial sessions, a final three day presentations session and individual consultations in between.

Dates and Location

  • Initial Sessions:
    • Wednesday, 23.10.24, 2pm to 4pm (Location: COS INF231, kHS (small lecture hall; 6231EG001))
    • Wednesday, 30.10.24, 2pm to 4pm (Location: COS INF231, kHS (small lecture hall; 6231EG001))
  • Individual Consultations Hours
  • Mandatory Consultations: 06. - 17.01., please get in touch via mail to set up a meeting.
  • Slide Check: Please send your preliminary slides via mail until 17.01.
  • Final Presentations: 20. - 24.01, 2-3 days (Location: TBA)

Prerequisites

Due to its research focus, the seminar requires some prior knowledge in mathematical statistics. Accordingly, the main target group are 3rd year B.Sc. Mathematics students and students in M.Sc. Mathematics or M.Sc. Data Science with appropriate course work. Required courses:
  • Introduction to Probability and Statistics
  • Mathematical Statistics 1

Grading:

The seminar is required to be graded. Details will be discussed in the initial seminar sessions.

Registration

If you are interested, please sign up for the class in MUESLI. Additionally, please sign up for the seminar in HeiCo (as soon as it is possible; we will send a Moodle reminder).

Moodle

The page on Moodle will be made accessible prior to the first seminar session. There you will find a list of research papers to choose from for your presentation.