Scaling knowledge graph embedding models for link prediction
- 5 April 2022
- conference paper
- conference paper
- Published by Association for Computing Machinery (ACM) in Proceedings of the 2nd European Workshop on Machine Learning and Systems
Abstract
Developing scalable solutions for training Graph Neural Networks (GNNs) for link prediction tasks is challenging due to the inherent data dependencies which entail high computational costs and a huge memory footprint. We propose a new method for scaling training of knowledge graph embedding models for link prediction to address these challenges. Towards this end, we propose the following algorithmic strategies: self-sufficient partitions, constraint-based negative sampling, and edge mini-batch training. The experimental evaluation shows that our scaling solution for GNN-based knowledge graph embedding models achieves a 16x speed up on benchmark datasets while maintaining a comparable model performance to non-distributed methods on standard metrics.Keywords
This publication has 15 references indexed in Scilit:
- Microsoft Academic Graph: When experts are not enoughQuantitative Science Studies, 2020
- Scalable Edge PartitioningPublished by Society for Industrial & Applied Mathematics (SIAM) ,2019
- Long-tail Relation Extraction via Knowledge Graph Embeddings and Graph Convolution NetworksPublished by Association for Computational Linguistics (ACL) ,2019
- Cross-lingual Knowledge Graph Alignment via Graph Convolutional NetworksPublished by Association for Computational Linguistics (ACL) ,2018
- Knowledge Graph Embedding: A Survey of Approaches and ApplicationsIEEE Transactions on Knowledge and Data Engineering, 2017
- Graph Edge Partitioning via Neighborhood HeuristicPublished by Association for Computing Machinery (ACM) ,2017
- An End-to-End Model for Question Answering over Knowledge Base with Cross-Attention Combining Global KnowledgePublished by Association for Computational Linguistics (ACL) ,2017
- GraphGenProceedings of the VLDB Endowment, 2015
- One trillion edgesProceedings of the VLDB Endowment, 2015
- Think Locally, Act Globally: Highly Balanced Graph PartitioningLecture Notes in Computer Science, 2013