Research map / Graph learning / Knowledge and temporal graphs
Learning to Extrapolate Knowledge: Transductive Few-shot Out-of-Graph Link Prediction
Jinheon Baek, Dong Bok Lee, Sung Ju Hwang
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 transductive message passing between unseen entities | I-GEN, yielding T-GEN · FB15k-237 and NELL-995 OOG link prediction | MRR: .348 -> .367 (FB15k-237 1-shot); .367 -> .382 (FB15k-237 3-shot); .278 -> .282 (NELL-995 1-shot); .285 -> .291 (NELL-995 | Improved |
| Add transductive message passing between unseen entities | I-GEN, yielding T-GEN · DeepDDI and BIOSNAP-sub 3-shot OOG | ROC: .946 -> .954 (DeepDDI); .608 -> .625 (BIOSNAP-sub) | Improved |
| Add transductive message passing between unseen entities | I-GEN, yielding T-GEN · DeepDDI and BIOSNAP-sub 3-shot OOG | PR: .681 -> .708 (DeepDDI); .062 -> .067 (BIOSNAP-sub) | Improved |
| Add transductive message passing between unseen entities | I-GEN, yielding T-GEN · DeepDDI and BIOSNAP-sub 3-shot OOG | Acc: .807 -> .815 (DeepDDI); .073 -> .089 (BIOSNAP-sub) | Improved |
| Add stochastic inference to the transductive layer | Deterministic T-GEN · Unseen-to-unseen link prediction | MRR | Improved |
| Remove many-shot-to-few-shot transfer strategy (ablation) | I-GEN · FB15k-237 and NELL-995 1- and 3-shot OOG | MRR: .348 -> .344 and .367 -> .362 (FB15k-237); .278 -> .272 and .285 -> .277 (NELL-995) | Hurt |
| Remove many-shot-to-few-shot transfer strategy (ablation) | T-GEN · FB15k-237 and NELL-995 1- and 3-shot OOG | MRR: .367 -> .362 and .382 -> .381 (FB15k-237); .282 -> .273 and .291 -> .290 (NELL-995) | Hurt |
| Use random rather than pre-trained DistMult embeddings (ablation) | I-GEN · FB15k-237 1- and 3-shot OOG | MRR: .348 -> .309; .367 -> .319 | Hurt |
| Use random rather than pre-trained DistMult embeddings (ablation) | T-GEN · FB15k-237 1- and 3-shot OOG | MRR: .367 -> .349; .382 -> .360 | Hurt |
| Replace DistMult initialization and score function with TransE (ablation) | I-GEN and T-GEN · FB15k-237 OOG link prediction | MRR | No clear effect |
| Train on 1-shot cases and test on 3-shot cases (ablation) | GEN · FB15k-237 and NELL-995 OOG link prediction | MRR | Improved |
Closest papers by meaning
- Convolutional 2D Knowledge Graph Embeddings (AAAI 2018 · similarity 0.88)
- How Does Knowledge Graph Embedding Extrapolate to Unseen Data: A Semantic Evidence View (AAAI 2022 · similarity 0.87)
- A Theory of Link Prediction via Relational Weisfeiler-Leman on Knowledge Graphs (NeurIPS 2023 · similarity 0.87)
- Embedding Logical Queries on Knowledge Graphs (NeurIPS 2018 · similarity 0.87)
- Toward a Unified Statistical Theory of Unsupervised Pretraining and Supervised Neural Knowledge Graph Learning (NeurIPS 2026 · similarity 0.86)
- Towards Global-Topology Relation Graph for Inductive Knowledge Graph Completion (AAAI 2025 · similarity 0.86)
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.
