AI Paper Atlas

Research map / Graph learning / GNN architectures

On the Bottleneck of Graph Neural Networks and its Practical Implications

Uri Alon, Eran Yahav

ICLR 2021 · Graph learning › GNN architectures · approach: GNN generalization theory · 89 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
Add a fully-adjacent final layerR-GIN · QM9average error: -39.54%Improved
Add a fully-adjacent final layerR-GAT · QM9average error: -44.58%Improved
Add a fully-adjacent final layerGGNN · QM9average error: -47.42%Improved
Add a fully-adjacent final layerGIN · NCI1error: 1.5%Improved
Add a fully-adjacent final layerGIN · ENZYMESerror: 8.1% over GIN-base; 2.5% over No StructImproved
Add a fully-adjacent final layerTuned GNNs · VarMisuse SeenProjectTestaccuracy: 88.4%Improved
Add a fully-adjacent final layerTuned GNNs · VarMisuse UnseenProjectTestaccuracy: 83.8%Improved
Double the hidden dimension instead of adding an FA layer (ablation)GCN · QM9average error: -5.50%Improved
Convert every layer to fully adjacent, removing reliance on original graph topology (ablation)GCN · QM9average error: +1520%Hurt
Add a second fully-adjacent layer (ablation)GCN + FA · QM9average error: -43.30% versus base; +FA is -43.40%No clear effect
Move the fully-adjacent layer to the penultimate position (ablation)GCN · QM9average error: -45.2%Improved
Use only 0.25 of the possible edges in the FA layer (ablation)GCN · QM9average error: -8.4%Improved

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