AI Paper Atlas

Research map / Graph learning / GNN architectures

Error-Driven Graph Augmentation for Mesh-Based PDE Surrogates

Xuan Minh Vuong Nguyen, Nissrine Akkari, Fabien Casenave, Jonathan Viquerat, Elie Hachem

ICML 2026 · Graph learning › GNN architectures · approach: Structural GNN variants

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
use error-guided augmentation instead of best non-error-guided equal-budget edges (ablation)MiSe-GNN · 2D profilemacro RRMSE: 0.0368 -> 0.0295Improved
use error-guided augmentation instead of best equal-budget control (ablation)MiSe-GNN · 2D MultiScHypElmacro RRMSE: 0.0242 -> 0.0224Improved
scale capacity from 10 MP/32 to 15 MP/80 (ablation)MiSe-GNN · 2D profilemacro RRMSE: 0.0431 -> 0.0372Improved
add DID input features (ablation)MiSe-GNN (15/80) · 2D profilemacro RRMSE: 0.0372 -> 0.0315Improved
add Sinus and Sph encodings to DID (ablation)MiSe-GNN (15/80) · 2D profilemacro RRMSE: 0.0315 -> 0.0295Improved
add Sinus and Sph encodings to DID (ablation)MiSe-GNN (15/80) · 2D profilePressure RRMSE: 0.02123 -> 0.02543Hurt
add OurFeat features to DID (ablation)MiSe-GNN (15/80) · 2D MultiScHypElmacro RRMSE: 0.0251 -> 0.0224Improved
give MGN and MGN Tree the OurFeat+DID features (stronger inputs do not close the gap) (ablation)MGN / MGN Tree · 2D MultiScHypElmacro RRMSE: 0.0269 and 0.0261 vs 0.0224 for MiSe-GNNNo clear effect
add hierarchical tree rewiring to MGN (ablation)MGN · 2D profile (DID+Sinus+Sph)macro RRMSE: 0.0356 -> 0.0376Hurt
add error-guided augmentation to MGNMGN · 2D MultiScHypEl (10/32)u1 RRMSE: 0.0419 -> 0.0185Improved
add error-guided augmentation to MGNMGN · 2D profile (10/32)Pressure RRMSE: 0.0624 -> 0.0290Improved

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