Research map / Graph learning / Knowledge and temporal graphs
Comprehensive Analysis of Negative Sampling in Knowledge Graph Representation Learning
Hidetaka Kamigaito, Katsuhiko Hayashi
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add Uniq subsampling (ablation) | HAKE · FB15k-237 | MRR: 32.3 -> 35.4 | Improved |
| Add Freq subsampling (ablation) | HAKE · FB15k-237 | MRR: 32.3 -> 34.9 | Improved |
| Replace Base subsampling with Uniq subsampling (ablation) | HAKE · FB15k-237 | MRR: 34.5 -> 35.4 | Improved |
| Add Uniq subsampling (ablation) | HAKE · WN18RR | MRR: 49.1 -> 49.8 | Improved |
| Replace Base subsampling with Uniq subsampling (ablation) | HAKE · WN18RR | MRR: 49.8 -> 49.8 | No clear effect |
| Add Uniq subsampling (ablation) | HAKE · YAGO3-10 | MRR: 53.4 -> 55.0 | Improved |
| Replace Base subsampling with Freq subsampling (ablation) | HAKE · YAGO3-10 | MRR: 54.3 -> 54.0 | Hurt |
| Increase the margin for TransE and RotatE (ablation) | Distance-based KGE models · FB15k-237 and WN18RR | MRR | Improved |
| Set a non-zero margin (ablation) | RESCAL, ComplEx, and DistMult · FB15k-237 and WN18RR | MRR | No clear effect |
| Increase negative sample count (ablation) | TransE and RotatE with original NS loss · FB15k-237 | MRR | Improved |
| Increase negative sample count (ablation) | RESCAL, ComplEx, and DistMult with original NS loss · FB15k-237 and WN18RR | MRR | No clear effect |
| Use the NS loss normalized by the number of negative samples (ablation) | KGE models · FB15k-237 and WN18RR | MRR | Mixed |
Closest papers by meaning
- DANS-KGC: Diffusion Based Adaptive Negative Sampling for Knowledge Graph Completion (AAAI 2026 · similarity 0.90)
- Does GCL Need a Large Number of Negative Samples? Enhancing Graph Contrastive Learning with Effective and Efficient Negative Sampling (AAAI 2025 · similarity 0.85)
- KGCRR: An Effective Metric-Driven Knowledge Graph Completion Framework by Designing a Novel Upper Bound Function with Adaptive Approximation to Reciprocal Rank (AAAI 2025 · similarity 0.84)
- KGDM: A Diffusion Model to Capture Multiple Relation Semantics for Knowledge Graph Embedding (AAAI 2024 · similarity 0.84)
- Neighborhood-Aware Negative Sampling for Student Knowledge and Behavior Modeling (AAAI 2025 · similarity 0.84)
- Towards Continual Knowledge Graph Embedding via Incremental Distillation (AAAI 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
