Diffusion Minimization and Sheaf Neural Networks for Recommender Systems
Keywords: Graph Neural Networks, Sheaves, Oversmoothing
TL;DR: Novel approach to minimise oversmothing via sheaves on graphs
Abstract: Graph Neural Networks (GNN) are well-known for successful applications in recommender systems. Despite recent advances in GNN development, various authors report that in certain cases GNN suffer from so-called oversmoothing problems. Sheaf Neural Networks (SNN) is one of the ways to address the issue of oversmoothing. In the present work we propose a novel approach for training SNN together with user and item embeddings. In that approach parameters of the sheaf are inferred via minimization of the classical BPR loss and sheaf diffusion on graphs subjected to orthogonality and consistency constraints. Performance of the novel technique is evaluated on synthetic test cases and standard benchmarks for recommendations.
Supplementary Material: zip
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: 7730
Loading