Topological Graph Neural Networks
Автор: Bastian Grossenbacher-Rieck
Загружено: 2021-11-18
Просмотров: 1763
This talk, similar to the one given @aatrn1, deals with a recent preprint of ours (https://arxiv.org/abs/2102.07835), in which we show that (1) topological features are more expressive than a standard Weisfeiler–Lehman test, and (2) they can help improve predictive performance in tasks that require structural information.
This talk was given in the Graph Representation Learning Reading Group at Mila (https://grlmila.github.io). Thanks a lot for the invitation!
Доступные форматы для скачивания:
Скачать видео mp4
-
Информация по загрузке: