Research map / Probabilistic and scientific modeling / Molecules and proteins
Complex Embeddings for Simple Link Prediction
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use complex embeddings with the Hermitian scoring function instead of real DistMult embeddings | Link prediction · FB15K | filtered MRR: 0.654 -> 0.692 | Helped |
| Use complex embeddings with the Hermitian scoring function instead of HolE | Link prediction · FB15K | filtered MRR: 0.524 -> 0.692 | Helped |
| Use complex embeddings with the Hermitian scoring function instead of HolE | Link prediction · WN18 | filtered MRR: 0.938 -> 0.941 | Helped |
| Increase negative samples per positive training triple from 1 to 10 (ablation) | ComplEx · FB15K | filtered MRR: up to +0.08 | Helped |
| Increase negative samples per positive training triple to 100 (ablation) | ComplEx · FB15K | filtered MRR: 0.737 | Helped |
| Increase negative samples per positive training triple from 100 to 200 (ablation) | ComplEx · FB15K | filtered MRR | Hurt |
| Increase negative samples per positive triple up to 50 (ablation) | ComplEx · FB15K | training time to convergence | Mixed |
| Use logistic negative log-likelihood instead of pairwise ranking loss (ablation) | Link prediction models · FB15K | performance | Helped |
| Vary batch size (ablation) | ComplEx · FB15K and WN18 | performance | No clear effect |
| Tune L2 regularization from no regularization to the optimal value (ablation) | Link prediction models · FB15K and WN18 | filtered MRR: up to +0.05 | Helped |
| Tune the initial learning rate (ablation) | ComplEx · FB15K | filtered MRR | Helped |
| Tune the initial learning rate (ablation) | ComplEx · WN18 | filtered MRR | No clear effect |
Closest papers by meaning
- SimplE Embedding for Link Prediction in Knowledge Graphs (NeurIPS 2018 · similarity 0.89)
- Multi-relational Poincaré Graph Embeddings (NeurIPS 2019 · similarity 0.87)
- On the Theoretical Limitations of Embedding-based Link Prediction (ICML 2026 · similarity 0.87)
- On Multi-Relational Link Prediction With Bilinear Models (AAAI 2018 · similarity 0.87)
- Bridging Theory and Practice in Link Representation with Graph Neural Networks (NeurIPS 2025 · similarity 0.85)
- Link Prediction in Multilayer Networks via Cross-Network Embedding (AAAI 2024 · similarity 0.85)
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.