Swedish web site
Geometric Deep Learning2026HT, 6.0 credits
|
|
Course plan
No of lectures
See organization
Recommended for
PhD students
The course was last given
N/A
Goals
The course aims to provide doctoral students with a solid understanding of the theoretical foundations and practical methods of Geometric Deep Learning. Participants will learn how deep learning models can be defined on non‑Euclidean domains such as graphs and manifolds, critically analyse recent research in the field, and connect course concepts to their own research problems.
Prerequisites
- Basic knowledge of machine learning and deep learning, including neural
networks and optimisation
- Familiarity with linear algebra, multivariable calculus, and probability
theory
- Practical experience in Python and at least one deep‑learning framework
(e.g. PyTorch or TensorFlow)
Organization
The course is organised as self‑study supported by 10–12 seminars. Participants prepare independently and discuss assigned chapters of the course literature and related resources during the seminars.
Content
- Independent reading and preparation
- Weekly or biweekly seminars with student‑led presentations
- (Optional) Individual project work connecting course topics to ongoing
research
Literature
The course is based on the materials at https://geometricdeeplearning.com/ and the accompanying textbook by Bronstein et al.: https://arxiv.org/abs/2104.13478.
Lectures
The course is based on seminars and self‑study.
Examination
Examination for 4 HP consists of:
- Two seminar presentations during the course (in small groups, based on weekly
topics)
- Attendance at a minimum of 80% of the seminars
Participants may obtain an additional 2 HP by completing a course‑related
project, including:
- A short written report (maximum 6 pages)
- Optional presentation at the final seminar
Examiner and course leader
Course leader: Pavlo Melnyk
Examiner: Fredrik Lindsten
Credits
4 HP, with an additional 2 HP available through an optional course project
Comments
Page responsible: Anne Moe
