AI Paper Atlas

Research map / Graph learning / GNN architectures

How hard is to distinguish graphs with graph neural networks?

Andreas Loukas

NeurIPS 2020 · Graph learning › GNN architectures · approach: Graph rewiring and scalable message passing · 17 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
Increase communication capacityMPNN graph-isomorphism classifiers · 12 graph and tree isomorphism taskstest accuracyImproved
Increase graph size from 6 to 12 nodesMPNN graph-isomorphism classifiers · X_graph^6, X_graph^8, X_graph^10, X_graph^12capacity required for 99% accuracy: 4, 8, 24, and 112Mixed
Use sufficient communication capacityMPNN graph-isomorphism classifiers · Graph and tree isomorphism tasksclassification accuracyImproved
Keep capacity below the theoretical lower-bound regionMPNN graph-isomorphism classifiers · Graph and tree isomorphism taskstest accuracyHurt
Increase depth and width while keeping capacity comparable (ablation)MPNN graph-isomorphism classifiers · Graph and tree isomorphism taskstest accuracyNo clear effect
Make the network deep enough for its receptive field to cover the graph (ablation)MPNN graph-isomorphism classifiers · Graph and tree isomorphism taskstest accuracyImproved
Use a large MPNN with one-hot node-order featuresGIN0 graph-isomorphism classifier · X_tree^22 testaccuracy: 82.92%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