AI Paper Atlas

Research map / Graph learning / GNN architectures

FlowX: Towards Explainable Graph Neural Networks via Message Flows

Shurui Gui, Hao Yuan, Jie Wang, Qicheng Lao, Kang Li, Shuiwang Ji

TPAMI 2023 · Graph learning › GNN architectures · approach: GNN explainability · 32 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 necessity-oriented coalition-weight training (FlowX_nec) to FlowX samplingFlowX · 9 datasets, GINFidelity+Improved
add sufficiency-oriented coalition-weight training (FlowX_suf) to FlowX samplingFlowX · 9 datasets, GINFidelity-Improved
use FlowX_suf instead of RC-ExplainerGraph-SST2 · Graph-SST2, GINFidelity-Hurt
train flow-level mask (FlowMask) instead of edge-level mask (GNNExplainer), same MI objective (ablation)GNNExplainer · 9 datasets, necessary explanations (Fidelity+)Fidelity+: better on 7 out of 9 datasetsImproved
train flow-level mask (FlowMask) instead of edge-level mask (GNNExplainer) (ablation)GNNExplainer · 9 datasets, sufficient explanations (Fidelity-)Fidelity-: better on 6 out of 9 datasetsImproved
use flow explanations instead of edge explanations (ablation)explainers · BA-Trafficfull recall sparsityImproved
use node-based explainers converted to edges instead of edge/flow onesGradCAM, DeepLIFT · BA-Trafficfull recall sparsityHurt
learn coalition weights on top of sampled Shapley scores (ablation)FlowX · explanation trainingexplanation qualityImproved

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