Abstract:The traditional model training approach based on negative sampling randomly samples a portion of negative samples for training, which can easily overlook important negative samples and adversely affect the training of knowledge graph embedding models. Some researchers have explored non-sampling model training frameworks that use all unobserved triples as negative samples to improve model training performance. However, both training methods inevitably introduce false negative samples and easy-to-separate negative samples that are far from the model's decision boundary, and they do not consider the adverse effects of long-tail entities and relations during training, thus limiting the improvement of model training performance. To address this issue, we propose a universal knowledge graph embedding framework based on high-quality negative sampling and weighting, called HNSW-KGE. First, we conduct pre-training based on the NS-KGE non-sampling training framework to quickly obtain an initial set of relatively high-quality embedding vector representations for all entities and relations. Second, we design a candidate negative sample set construction strategy that samples a certain number of negative samples that are neither false negatives nor easy-to-separate negatives for all positive triples, based on the embedding vectors obtained from pre-training. This ensures the provision of high-quality negative samples for model training. Finally, we apply weighting to the loss function based on the frequency of the entities and relations appearing in the triples to mitigate the adverse effects of long-tail entities and relations on model training. Experiments conducted on benchmark datasets FB15K237 and WN18RR using various knowledge graph embedding models demonstrate that our proposed framework HNSW-KGE, based on high-quality negative sampling and weighting, achieves better training performance and exhibits versatility, making it applicable to various types of knowledge embedding models.

Universal Knowledge Graph Embedding Framework Based on High-Quality Negative Sampling and Weighting

Efficient Non-Sampling Knowledge Graph Embedding

Entity Similarity-Based Negative Sampling for Knowledge Graph Embedding

Knowledge Graph Embedding Based on Adaptive Negative Sampling.

Efficient, Simple and Automated Negative Sampling for Knowledge Graph Embedding

Simple and Automated Negative Sampling for Knowledge Graph Embedding

NSCaching: Simple and Efficient Negative Sampling for Knowledge Graph Embedding

A Novel Negative Sampling Based on Frequency of Relational Association Entities for Knowledge Graph Embedding

Diversified and Adaptive Negative Sampling on Knowledge Graphs

Negative Sampling with Adaptive Denoising Mixup for Knowledge Graph Embedding

A Novel Negative Sample Generating Method for Knowledge Graph Embedding

Entity Aware Negative Sampling with Auxiliary Loss of False Negative Prediction for Knowledge Graph Embedding

Hierarchical Type Enhanced Negative Sampling for Knowledge Graph Embedding

Weighted Knowledge Graph Embedding

Unified Interpretation of Smoothing Methods for Negative Sampling Loss Functions in Knowledge Graph Embedding

Negative Sampling in Knowledge Graph Representation Learning: A Review

Incorporating GAN for Negative Sampling in Knowledge Representation Learning

TuckerDNCaching: high-quality negative sampling with tucker decomposition

Modality-Aware Negative Sampling for Multi-modal Knowledge Graph Embedding

Network embedding based on high-degree penalty and adaptive negative sampling

HNS: Hierarchical Negative Sampling for Network Representation Learning