NICE: Scale-Stable Perturbations for Graph Neural Network Explanations via Noise Corruption
2026-08-17 • Machine Learning
Machine LearningArtificial Intelligence
AI summaryⓘ
The authors studied how explanations for Graph Neural Networks (GNNs) are made by changing parts of the input graph and seeing how predictions change. They found that the common way of doing this, called Element-wise Masking, causes a problem they call Scale Drift, where information gets distorted as it passes through layers. To fix this, they created a new method called Noise Corruption that keeps the scale stable while still changing the messages in the graph. Using this method, their explanation framework NICE gives more reliable and clearer reasons for the model's decisions, as confirmed by experiments.
Graph Neural NetworkPost-hoc ExplanationPerturb-Query ParadigmElement-wise MaskingScale DriftNoise CorruptionMessage PassingModel FaithfulnessStochastic Restoration BoundaryIntegrated Gradient
Authors
Ziluowen Luo, Jun Yin, Ruochen Liu, Ming Cheng, Shirui Pan, Chengqi Zhang, Senzhang Wang
Abstract
Post-hoc Graph Neural Network (GNN) explainers commonly follow a Perturb-Query paradigm, inferring the importance of graph elements based on queried predictions to perturbed inputs. However, such perturbations often introduce substantial distribution shift, undermining the reliability of the queried predictions used to derive explanations. While existing efforts mainly improve perturbed graphs or stabilize model predictions on them, we revisit the perturbation mechanism itself. We show that the widely used Element-wise Masking(EM) suppresses edge-induced messages toward zero, causing deterministic scale contraction that accumulates across message-passing layers, a phenomenon we term Scale Drift. Consequently, prediction changes under EM may conflate information corruption with deviations in propagation scale. As a scale-stable alternative to EM, we introduce Noise Corruption (NC), which perturbs each message through matched-norm random-direction corruption while preserving the expected squared message norm. Building on NC, we propose NICE, a Noise Corruption-based explanation framework, which learns a Stochastic Restoration Boundary (SRB) under NC-induced uncertainty, balancing target-prediction restoration against compactness. Furthermore, Boundary-Integrated Gradient (BIG) converts this boundary into edge attributions by accumulating each edge's contribution to reducing restoration risk along the restoration path. Experiments across multiple benchmarks demonstrate stronger explanation performance and model faithfulness while confirming that NC substantially reduces the Scale Drift induced by masking.