Simplicial Representation Learning with Neural k-Forms

Geometric deep learning extends deep learning to incorporate information about the geometry and topology of data, especially in complex domains like graphs. Many existing methods in this field rely on message passing. However in this talk we will take a new approach by combining the theory of differential k-forms in euclidean space with the geometric information of graphs and complexes given by embeddings of node coordinates. This method offers interpretability and geometric consistency without the use of message passing.