AI Paper Atlas

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

NeurIPS 2022 · Graph learning › GNN architectures · approach: Expressive equivariant message passing · 7 citations (OpenAlex)

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 changedWhereMeasuredEffect
test on 10x larger graph (OOD) versus in-distribution inductiveGraphSAGE · SBM N_tr=1e3mcc (%): 19.34 -> -6.56Hurt
test on 10x larger graph (OOD) versus in-distribution inductiveGIN · SBM N_tr=1e3Hit@100 (%): 43.90 -> 0.00Hurt
test on 10x larger graph (OOD) versus in-distribution inductivegMPNN** (learn Psi) · SBM N_tr=1e3mcc (%): 94.23 -> 93.94No clear effect
use pairwise embedding gMPNN** instead of node embedding GCNGCN* · SBM N_tr=1e3 OODmcc (%): -4.88 -> 93.94Improved
use pairwise embedding gMPNN** instead of node embedding GraphSAGEGraphSAGE* · SBM N_tr=1e3 OODbalanced accuracy (%): 49.32 -> 96.88Improved
learn Psi with an MLP instead of fixed PsigMPNN** · SBM N_tr=1e3 OODHit@10 (%): 96.74 -> 96.97No clear effect
learn Psi with an MLP instead of fixed PsigMPNN** · ogbl-ddi OODHit@10 (%): 9.31 -> 57.97Improved
learn Psi with an MLP instead of fixed PsigMPNN** · ogbl-ddi OODmcc (%): 75.42 -> 72.04Hurt
use pairwise gMPNN** instead of node embeddingsGCN* · ogbl-ddi OODmcc (%): 55.03 -> 75.42Improved
enlarge training graph from 1e3 to 1e4 nodesGCN* · SBM OODmcc (%): -4.88 -> 68.93Mixed
move from transductive to inductive evaluationGraphSAGE* · SBM N_tr=1e3mcc (%): 95.42 -> 19.34Hurt
move from transductive to inductive evaluationgMPNN** (learn Psi) · SBM N_tr=1e3mcc (%): 94.23 -> 94.23No clear effect

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy