Low Variance: A Bottleneck in Diffusion-Based Graph Imputation

Daeho Um, Sunoh Kim, Jiwoong Park, Jongin Lim, Seong Jin Ahn, Seulki Park

Keywords: diffusion-based imputation, missing features, graph neural networks

TL;DR: For graphs with missing features, we identify a problem in existing diffusion methods and propose a novel scheme that addresses this issue.

Abstract: In this paper, we tackle learning tasks on graphs with missing features, improving the applicability of graph neural networks to real-world graph-structured data. Existing imputation methods based upon graph diffusion produce channels that have nearly identical values within each channel, and these low-variance channels contribute very little to performance in graph learning tasks. To prevent diffusion-based imputation from producing low-variance channels, we introduce synthetic features that address the cause of the production, thereby increasing variance in low-variance channels. Since the synthetic features prevent diffusion-based imputation models from generating meaningless feature values shared across all nodes, our synthetic feature propagation design prevents significant performance degradation, even under extreme missing rates. Extensive experiments demonstrate the effectiveness of our scheme across various graph learning tasks with missing features, ranging from low to extremely high missing rates. Moreover, we provide empirical evidence and theoretical proof that validate the low-variance problem.

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: 8202