COINs: Model-based Accelerated Inference for Knowledge Graphs
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.
Keywords: Knowledge Graph Inference, Scalability, Graph Embeddings, Community Structure
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
TL;DR: Community structure leveraged for accelerating embeddings-based knowledge graph inference while preserving prediction power
Abstract: We introduce **CO**mmunity **IN**formed graph embedding**s** (COINs), for accelerating link prediction and query answering models for knowledge graphs. COINs employ a community-detection-based graph data augmentation procedure, followed by a two-step prediction pipeline: node localization via community prediction and then localization within the predicted community. We describe theoretically justified criteria for gauging the applicability of our approach in our setting with a direct formulation of the reduction in time complexity. Additionally, we provide numerical evidence of superior scalability in model evaluation cost (average reduction factor of 6.413 $\pm$ 3.3587 on a single-CPU-GPU machine) with admissible effects on prediction performance (relative error to baseline 0.2389 $\pm$ 0.3167 on average).
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: 6340
Loading