AI Paper Atlas

Research map / Graph learning / Graph self-supervised learning

Evaluating Robustness and Uncertainty of Graph Models Under Structural Distributional Shifts

Gleb Bazhenov, Denis Kuznedelev, Andrey Malinin, Artem Babenko, Liudmila Prokhorenkova

NeurIPS 2023 · Graph learning › Graph self-supervised learning · approach: Graph out-of-distribution generalization · 1 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 ERM on nodes selected by locality rather than popularityERM · Structural shifts, average across datasetsAccuracy drop (ID to OOD): -13.56\,\% locality vs -4.39\,\% popularityHurt
Test SE on density-based rather than popularity-based shiftsSE · Structural shifts, average across datasetsAUROC: 81.40 vs 52.10Hurt
Use Mixup instead of ERMGraph node classification · Popularity and locality structural shiftsAccuracy rank on OOD nodes: 2.1 (Popularity); 2.4 (Locality)Improved
Use DE instead of other evaluated methodsGraph node classification · Density structural shift, OOD nodesAccuracy rank: 2.0Improved
Use DE uncertainty estimates instead of other evaluated methodsOOD detection · Locality and density structural shiftsAUROC rank: 1.1 (Locality); 2.7 (Density)Improved
Add the modified architecture (MLP preprocessing, skip connections, fewer layers, and SAGE) to ERM (ablation)ERM · Popularity and density shifts, OOD nodesPerformance versus baseline win/tie/loss: 5/2/0Improved
Add the modified architecture (MLP preprocessing, skip connections, fewer layers, and SAGE) to ERM (ablation)ERM · Locality shift, OOD nodesPerformance versus baseline win/tie/loss: 3/2/2Mixed
Add the modified architecture (MLP preprocessing, skip connections, fewer layers, and SAGE) to SE (ablation)SE · Popularity and locality shiftsPerformance versus baseline win/tie/loss: 0/0/7 (Popularity); 0/1/6 (Locality)Hurt
Add the modified architecture (MLP preprocessing, skip connections, fewer layers, and SAGE) to SE (ablation)SE · Density shiftPerformance versus baseline win/tie/loss: 4/0/3Mixed
Increase the ID share from 70% to 90% (ablation)ERM · Locality shiftAverage accuracy drop (ID to OOD): -4.65\,\% -> -1.02\,\%Improved
Increase the ID share from 70% to 90% (ablation)SE · Locality shiftAverage AUROC: 78.71 -> 72.95Hurt

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