Naturality-Guided Hyperedge Disentanglement for Message Passing Hypergraph Neural Network
Keywords: Graph Neural Network, Hypergraph, Category Theory, Categorcial Deep Learning, Disentangled Representation Learning
TL;DR: Reflecting the interaction context of each hyperedge is important for real world problems. We propose a simple but effective hyperedge disentangling model that can reflect inherent heterogenity (hyperedge type) even in homogeneous data
Abstract: Hypergraph data structure has been widely used to store information or meaning derived from group interactions, meaning that each hyperedge inherently contains the context of their interactions. For example, a set of genes or a genetic pathway can be represented as a hyperedge to express the interaction of multiple genes that collaboratively perform a biological function (i.e., interaction context). However, most existing hypergraph neural networks cannot reflect the interaction context of each hyperedge due to their limited capability in capturing important or relevant factors therein. In this paper, we propose a \textbf{simple but effective} hyperedge disentangling method, \textbf{Natural-HNN}, that captures inherent hyperedge types or the interaction context of an hyperedge. We devised a novel guidance for hyperedge disentanglement based on the naturality condition in the category theory. In our experiments, we applied our model to hypergraphs of genetic pathways for the cancer subtype classification task, and showed that our model outperforms baselines by capturing the functional semantic similarity of genetic pathways.
Primary Area: learning on graphs and other geometries & topologies
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 13571
Loading