Research map / Graph learning / Knowledge and temporal graphs
RotatE: Knowledge Graph Embedding by Relational Rotation in Complex Space
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-space relational rotations instead of modulus-constrained phase-only embeddings (ablation) | RotatE · FB15k-237 | MRR: .328 -> .338 | Helped |
| Use complex-space relational rotations instead of modulus-constrained phase-only embeddings (ablation) | RotatE · WN18RR | MRR: .462 -> .476 | Helped |
| Use complex-space relational rotations instead of modulus-constrained phase-only embeddings (ablation) | RotatE · FB15k | MRR: .799 -> .797 | No clear effect |
| Use self-adversarial rather than uniform negative sampling (ablation) | RotatE with negative sampling loss · FB15k-237 | MRR: .297 -> .338 | Helped |
| Use self-adversarial rather than uniform negative sampling (ablation) | TransE with negative sampling loss · FB15k-237 | MRR: .297 -> .332 | Helped |
| Replace negative sampling loss with margin-based ranking criterion (ablation) | RotatE with self-adversarial sampling · FB15k-237 | MRR: .338 -> .322 | Hurt |
| Replace negative sampling loss with margin-based ranking criterion (ablation) | RotatE with uniform sampling · FB15k-237 | MRR: .297 -> .293 | No clear effect |
| Use self-adversarial rather than uniform negative sampling (ablation) | TransE with margin-based ranking criterion · FB15k-237 | MRR: .306 -> .333 | Helped |
| Use self-adversarial rather than uniform negative sampling | TransE · WN18 | MRR: .433 -> .736 | Helped |
| Use self-adversarial rather than uniform negative sampling | TransE · WN18RR | MRR: .186 -> .223 | Helped |
| Use RotatE rather than ComplEx for composition-heavy Countries tasks | Link prediction models · Countries S3 | AUC-PR: 0.43\pm0.07 -> \mathbf{0.95\pm0.00} | Helped |
| Use TransE rather than RotatE | Link prediction models · Countries S3 | AUC-ROC: 0.95 -> 0.96 | Helped |
Closest papers by meaning
- Rot-Pro: Modeling Transitivity by Projection in Knowledge Graph Embedding (NeurIPS 2021 · similarity 0.90)
- Learning Hierarchy-Aware Knowledge Graph Embeddings for Link Prediction (AAAI 2020 · similarity 0.90)
- Dual Quaternion Knowledge Graph Embeddings (AAAI 2021 · similarity 0.89)
- 5* Knowledge Graph Embeddings with Projective Transformations (AAAI 2021 · similarity 0.89)
- ChronoR: Rotation Based Temporal Knowledge Graph Embedding (AAAI 2021 · similarity 0.89)
- Towards Global-Topology Relation Graph for Inductive Knowledge Graph Completion (AAAI 2025 · similarity 0.87)
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.