Research map / Graph learning / GNN architectures
OOD Link Prediction Generalization Capabilities of Message-Passing GNNs in Larger Test Graphs
Yangze Zhou, Gitta Kutyniok, Bruno Ribeiro
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 |
|---|---|---|---|
| test on 10x larger graph (OOD) versus in-distribution inductive | GraphSAGE · SBM N_tr=1e3 | mcc (%): 19.34 -> -6.56 | Hurt |
| test on 10x larger graph (OOD) versus in-distribution inductive | GIN · SBM N_tr=1e3 | Hit@100 (%): 43.90 -> 0.00 | Hurt |
| test on 10x larger graph (OOD) versus in-distribution inductive | gMPNN** (learn Psi) · SBM N_tr=1e3 | mcc (%): 94.23 -> 93.94 | No clear effect |
| use pairwise embedding gMPNN** instead of node embedding GCN | GCN* · SBM N_tr=1e3 OOD | mcc (%): -4.88 -> 93.94 | Improved |
| use pairwise embedding gMPNN** instead of node embedding GraphSAGE | GraphSAGE* · SBM N_tr=1e3 OOD | balanced accuracy (%): 49.32 -> 96.88 | Improved |
| learn Psi with an MLP instead of fixed Psi | gMPNN** · SBM N_tr=1e3 OOD | Hit@10 (%): 96.74 -> 96.97 | No clear effect |
| learn Psi with an MLP instead of fixed Psi | gMPNN** · ogbl-ddi OOD | Hit@10 (%): 9.31 -> 57.97 | Improved |
| learn Psi with an MLP instead of fixed Psi | gMPNN** · ogbl-ddi OOD | mcc (%): 75.42 -> 72.04 | Hurt |
| use pairwise gMPNN** instead of node embeddings | GCN* · ogbl-ddi OOD | mcc (%): 55.03 -> 75.42 | Improved |
| enlarge training graph from 1e3 to 1e4 nodes | GCN* · SBM OOD | mcc (%): -4.88 -> 68.93 | Mixed |
| move from transductive to inductive evaluation | GraphSAGE* · SBM N_tr=1e3 | mcc (%): 95.42 -> 19.34 | Hurt |
| move from transductive to inductive evaluation | gMPNN** (learn Psi) · SBM N_tr=1e3 | mcc (%): 94.23 -> 94.23 | No clear effect |
Closest papers by meaning
- A Topological Perspective on Demystifying GNN-Based Link Prediction Performance (ICLR 2024 · similarity 0.88)
- Learning to Reweight for Generalizable Graph Neural Network (AAAI 2024 · similarity 0.87)
- Structural Information Enhanced Graph Representation for Link Prediction (AAAI 2024 · similarity 0.87)
- Towards Dynamic Message Passing on Graphs (NeurIPS 2024 · similarity 0.87)
- GNNs Meet Sequence Models Along the Shortest-Path: an Expressive Method for Link Prediction (NeurIPS 2026 · similarity 0.87)
- Sign is Not a Remedy: Multiset-to-Multiset Message Passing for Learning on Heterophilic Graphs (ICML 2024 · 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.
