Organizers: Christian Koke, Abhishek Saroha

Please forward any queries related to the seminar to: graphlearning-ss23@vision.in.tum.de.

Assignment to the lab is done via the matching system. In addition to applying to the matching system please remember to also send in your application via e-mail at the latest by 14 February 2023.

The preliminary meeting will take place on 2nd February 2023 at 10 am via https://tum-conf.zoom.us/j/69836253691?pwd=cGl6R2FzdEt2RElydWlCMk40NEliZz09. We will post the slides here after the meeting in case you cannot attend, but still interested to apply. Attending the preliminary meeting is not obligatory, but highly encouraged. UPDATE: The preliminary meeting slides (ss23) have been uploaded.

The field of graph learning has recently witnessed a significant surge in interest with with emerging applications of networks on graph structured data ranging from topics in drug discovery over traffic control to guiding mathematicians intuition in the process of postulating new conjectures and and proving new theorems. This seminar focuses on recent research and emergent paradigms in graph learning. Possible presentation topics range from mathematical foundations to architectures tailored to specific applications such as e.g. molecular property predictions.

• Good knowledge of basic mathematics such as linear algebra, analysis, probability and numerics, graph theory etc. is required.

• Participation in at least one of the offered deep learning lectures at TUM is highly recommended. For e.g. 1, 2, 3 etc.

The seminar will take place in a block format on 4th and 5th April completely online via Zoom. After the kick-off meeting in early March, the students will be given to choose from a list of recent works in graph learning and present their findings as a short presentation during the seminar. Afterwards, they will be required to compile their results as a short report to be submitted towards the end of the semester.

The list of available papers can be found here (password protected).