DiRaGNN: Attention-Enhanced Entity Ranking for Sparse Graph Networks

Fiza Husain, Anjaly Parayil, Ayush Choure, Rujia Wang, Chetan Bansal

Keywords: Heterogeneous Graphs, Graph Neural Networks, Recommendation System

Abstract:

Sparsity in both the structural and engagement information presents a core challenge in entity ranking problems for graph networks. The interaction dynamics of entities are often characterized by limited structural and engagement information which results in inferior performance of the state-of-the-art approaches. In this work, we present DiRaGNN, an attention-enhanced entity ranking model designed
to address the problem of dimension recommendation and ranking for automated watchdogs in the cloud setting. DiRaGNN is inspired by transformer architectures and utilizes a multi-head attention mechanism to focus on heterogeneous neighbors and their attributes. Additionally, our model employs multi-faceted loss functions to optimize for relevant recommendations and reduce popularity bias. To manage computational complexity, we sample a local subgraph that includes multiple hops of neighbors. Empirical evaluations demonstrate significant improvements over existing methods, with our model achieving a 39.7% increase in MRR.

Primary Area: unsupervised, self-supervised, semi-supervised, and supervised representation learning

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: 13965