Research map / Graph learning / Knowledge and temporal graphs
GraphMSE: Efficient Meta-path Selection in Semantically Aligned Feature Space for Graph Neural Networks
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 GraphMSE's automatic meta-path selection and embedding rather than the strongest listed baseline | Node classification · ACM, 20% training nodes | Macro F1: 92.29±0.63 -> 92.97±0.38 | Helped |
| Use GraphMSE rather than Metapath2vec | Node classification · DBLP, 20% training nodes | Macro F1: 92.14±0.27 -> 94.03±0.22 | Helped |
| Use GraphMSE rather than GTN | Node classification · IMDB, 20% training nodes | Macro F1: 57.59±0.75 -> 57.78±1.90 | No clear effect |
| Use GraphMSE rather than GTN's matrix-multiplication extraction | DBLP · DBLP | training time to lowest validation error (s): 5527.91 -> 22.94 | Helped |
| Keep only top-k selected meta-paths | GraphMSE · IMDB, 20% training nodes | Macro F1: 57.78±1.90 -> 59.31±2.88 | Helped |
| Replace SUM aggregation with MEAN aggregation (ablation) | GraphMSE · IMDB, 20% training nodes | Macro F1: 58.46±2.56 -> 48.46±2.69 | Hurt |
| Add semantic feature space alignment (ablation) | GraphMSE · IMDB, 20% training nodes | Macro F1: 56.93±2.11 -> 58.46±2.56 | Helped |
| Preserve intermediate nodes in meta-path instances (ablation) | GraphMSE · IMDB, 20% training nodes | Macro F1: 52.30±2.69 -> 58.46±2.56 | Helped |
| Use semantic feature space alignment (ablation) | GraphMSE · IMDB meta-path attention weights | attention-weight variance: MD: 5.86±14.47 -> 19.33±4.72; MDM: 0.68±1.83 -> 13.26±1.83 | Helped |
| Increase neighborhood sampling limit excessively (λ > 10) (ablation) | GraphMSE · IMDB | node classification performance | Hurt |
| Use excessively long candidate meta-paths (ablation) | GraphMSE · Node classification across ACM, DBLP, and IMDB | node classification performance | Hurt |
Closest papers by meaning
- Defining and Discovering Hyper-meta-paths for Heterogeneous Hypergraphs (NeurIPS 2025 · similarity 0.87)
- Reinforcement Learning Based Meta-Path Discovery in Large-Scale Heterogeneous Information Networks (AAAI 2020 · similarity 0.87)
- Long-range Meta-path Search on Large-scale Heterogeneous Graphs (NeurIPS 2024 · similarity 0.87)
- Heterogeneous Graph Neural Network on Semantic Tree (AAAI 2025 · similarity 0.86)
- Graph Metanetworks for Processing Diverse Neural Architectures (ICLR 2024 · similarity 0.86)
- HeterSEED: Semantics–Structure Decoupling for Heterogeneous Graph Learning under Heterophily (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.