🤖 AI Summary
Existing negative sampling methods in knowledge graph embedding suffer from insufficient diversity and adaptability, yielding low-informative negative triples and limiting model discriminability. To address this, we propose DANS—a generative adversarial negative sampling framework. DANS employs a bidirectional generator with separate head- and tail-entity pathways to enhance the structural and semantic diversity of negative triples. Furthermore, it introduces a local adaptive parameterization mechanism operating at both entity- and relation-levels, enabling fine-grained, highly discriminative negative sample generation. This work establishes the first “global diversity + individual discriminability” co-design paradigm for negative sampling. Extensive experiments on FB15k-237, WN18RR, and YAGO3-10 demonstrate that DANS significantly improves link prediction performance, achieving an average 3.2% gain in Mean Reciprocal Rank (MRR). Qualitative analysis confirms that DANS-generated negatives exhibit superior semantic distinctiveness and relational plausibility compared to baselines.
📝 Abstract
In knowledge graph embedding, aside from positive triplets (ie: facts in the knowledge graph), the negative triplets used for training also have a direct influence on the model performance. In reality, since knowledge graphs are sparse and incomplete, negative triplets often lack explicit labels, and thus they are often obtained from various sampling strategies (eg: randomly replacing an entity in a positive triplet). An ideal sampled negative triplet should be informative enough to help the model train better. However, existing methods often ignore diversity and adaptiveness in their sampling process, which harms the informativeness of negative triplets. As such, we propose a generative adversarial approach called Diversified and Adaptive Negative Sampling DANS on knowledge graphs. DANS is equipped with a two-way generator that generates more diverse negative triplets through two pathways, and an adaptive mechanism that produces more fine-grained examples by localizing the global generator for different entities and relations. On the one hand, the two-way generator increase the overall informativeness with more diverse negative examples; on the other hand, the adaptive mechanism increases the individual sample-wise informativeness with more fine-grained sampling. Finally, we evaluate the performance of DANS on three benchmark knowledge graphs to demonstrate its effectiveness through quantitative and qualitative experiments.