Entity Aware Negative Sampling with Auxiliary Loss of False Negative Prediction for Knowledge Graph Embedding
This addresses the bottleneck of negative sample quality in knowledge graph embedding, which is crucial for downstream applications, representing an incremental improvement over existing negative sampling techniques.
The paper tackles the problem of generating high-quality negative samples for knowledge graph embedding by proposing Entity Aware Negative Sampling (EANS), which uses Gaussian distribution in entity index space to sample negative entities similar to positive ones, and includes an auxiliary loss for false negative prediction. The method outperforms state-of-the-art negative sampling methods on standard benchmarks and achieves competitive performance with only one negative sample.
Knowledge graph (KG) embedding is widely used in many downstream applications using KGs. Generally, since KGs contain only ground truth triples, it is necessary to construct arbitrary negative samples for representation learning of KGs. Recently, various methods for sampling high-quality negatives have been studied because the quality of negative triples has great effect on KG embedding. In this paper, we propose a novel method called Entity Aware Negative Sampling (EANS), which is able to sample negative entities resemble to positive one by adopting Gaussian distribution to the aligned entity index space. Additionally, we introduce auxiliary loss for false negative prediction that can alleviate the impact of the sampled false negative triples. The proposed method can generate high-quality negative samples regardless of negative sample size and effectively mitigate the influence of false negative samples. The experimental results on standard benchmarks show that our EANS outperforms existing the state-of-the-art methods of negative sampling on several knowledge graph embedding models. Moreover, the proposed method achieves competitive performance even when the number of negative samples is limited to only one.