Networks seminar
Welcome to the homepage of the Networks seminars, a weekly seminar series on networks, complex systems, and related topics held in the Mathematical Institute. In this year's series, we will alternate between regular talks and "fresh from the arXiv" talks (FFTA) in which we invite the author of a recently published (pre)print to discuss their work. Suggestions are always welcome!
The Networks seminar usually takes place on Tuesdays at 14:00-15:00. In line with current regulation, we are excited to announce that the seminars will now run with a new hybrid format that will allow attendees to choose whether to join our group in person in room C1 at the Mathematical Institute, or to attend remotely on Zoom. A link to the event will be made available in the schedule of upcoming talks below (for logged-in users) and via the mailing list.
To sign up to our mailing list simply send an empty email to the following address:
@email
If you would like to give a presentation at our seminar, please do not hesitate to contact the organisers Erik Hörmann and Yu Tian. The presentation can be either about your own work or on some (recent) interesting article on networks or on complex systems in general.
In case you missed any of the talks, we will also make recordings of the talks available on our youtube channel.
Upcoming Seminars
Permutation Equivariance in Graph Neural Controlled Differential Equations for Dynamic Graph Representation Learning
Abstract
Many systems in the natural sciences and beyond exhibit complex relational structure that changes over time. Social networks evolve as relationships change, traffic patterns vary throughout the day, and protein–protein interactions shift with cellular conditions. Learning these dynamics from data is a challenging problem. A recent approach in this area, Graph Neural Controlled Differential Equations, extends Neural CDEs from paths on Euclidean domains to paths on graph domains. In this talk, we discuss an extension of this framework that respects the geometry of the underlying set and is equivariant to permutations of the node ordering. We will discuss empirical advantages of this modification, as well as benefits of the formulation as a continuous-time model.
You can also find a list of all talks (with abstracts) prior to 2018 here, and the former website
of the Networks journal club at the Oxford complexity center (CABDyN) here.
Unsubscribing?
Simply send an empty email to
@email