Research map / Graph learning / Knowledge and temporal graphs
Knowledge Graph Alignment Network with Gated Multi-Hop Neighborhood Aggregation
Zequn Sun, Chengming Wang, Wei Hu, Muhao Chen, Jian Dai, Wei Zhang, Yuzhong Qu
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 neighborhood augmentation edges between pre-aligned counterparts (ablation) | AliNet · DBP15K (DBP_ZH-EN, DBP_JA-EN, DBP_FR-EN) and DWY100K (DBP-WD, DBP-YG) | Hits@1: 0.012-0.037 | Improved |
| Add relation loss (ablation) | AliNet · DBP15K and DWY100K | Hits@1: 0.539 -> 0.525; 0.549 -> 0.539; 0.552 -> 0.535; 0.690 -> 0.679; 0.786 -> 0.773 | Improved |
| Aggregate two-hop neighbors in addition to one-hop neighbors (ablation) | AliNet · DBP15K | Hits@1: 0.487 -> 0.539; 0.507 -> 0.549; 0.508 -> 0.552 | Improved |
| Replace gating with addition for combining hop representations (ablation) | AliNet · DBP15K | Hits@1: 0.539 -> 0.498; 0.549 -> 0.515; 0.552 -> 0.501 | Hurt |
| Replace proposed attention with GAT attention (ablation) | AliNet · DBP15K | Hits@1: 0.539 -> 0.517; 0.549 -> 0.531; 0.552 -> 0.523 | Hurt |
| Use GCN layers for mixed two-hop neighborhood aggregation (ablation) | AliNet · DBP15K | Hits@1: 0.539 -> 0.227; 0.549 -> 0.294; 0.552 -> 0.258 | Hurt |
| Increase hop aggregation depth from two to three (ablation) | AliNet · DBP15K | Hits@1: 0.539 -> 0.461; 0.549 -> 0.484; 0.552 -> 0.450 | Hurt |
| Increase hop aggregation depth from two to four (ablation) | AliNet · DBP15K | Hits@1: 0.539 -> 0.386; 0.549 -> 0.407; 0.552 -> 0.373 | Hurt |
| Use three or four network layers instead of two (ablation) | AliNet · DBP15K | Hits@1 | Hurt |
| Use combined representations from all layers rather than a layer-specific representation (ablation) | AliNet · DBP15K | Hits@1 | Improved |
| Use distant neighbors to compensate for non-isomorphic direct neighborhoods | AliNet · DBP15K | one-hop neighbor overlap coefficient | Improved |
Closest papers by meaning
- Logic Attention Based Neighborhood Aggregation for Inductive Knowledge Graph Embedding (AAAI 2019 · similarity 0.88)
- Dynamic Knowledge Graph Alignment (AAAI 2021 · similarity 0.88)
- Chaining 2-FWL GNNs for Combinatorial Graph Alignment (NeurIPS 2026 · similarity 0.85)
- Towards Global-Topology Relation Graph for Inductive Knowledge Graph Completion (AAAI 2025 · similarity 0.85)
- Heterogeneous Graph Neural Network on Semantic Tree (AAAI 2025 · similarity 0.85)
- A Unified Graph Language Model for Multi-Domain Multi-Task Graph Alignment Instruction Tuning (NeurIPS 2026 · 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.
