
Generative Causal Explanations for Graph Neural Networks
This paper presents Gem, a modelagnostic approach for providing interpr...
read it

Perturb More, Trap More: Understanding Behaviors of Graph Neural Networks
While graph neural networks (GNNs) have shown a great potential in vario...
read it

Parameterized Explainer for Graph Neural Network
Despite recent progress in Graph Neural Networks (GNNs), explaining pred...
read it

XGNN: Towards ModelLevel Explanations of Graph Neural Networks
Graphs neural networks (GNNs) learn node features by aggregating and com...
read it

FairGNN: Eliminating the Discrimination in Graph Neural Networks with Limited Sensitive Attribute Information
Graph neural networks (GNNs) have shown great power in modeling graph st...
read it

Efficient Probabilistic Logic Reasoning with Graph Neural Networks
Markov Logic Networks (MLNs), which elegantly combine logic rules and pr...
read it

Principal Neighbourhood Aggregation for Graph Nets
Graph Neural Networks (GNNs) have been shown to be effective models for ...
read it
PGMExplainer: Probabilistic Graphical Model Explanations for Graph Neural Networks
In Graph Neural Networks (GNNs), the graph structure is incorporated into the learning of node representations. This complex structure makes explaining GNNs' predictions become much more challenging. In this paper, we propose PGMExplainer, a Probabilistic Graphical Model (PGM) modelagnostic explainer for GNNs. Given a prediction to be explained, PGMExplainer identifies crucial graph components and generates an explanation in form of a PGM approximating that prediction. Different from existing explainers for GNNs where the explanations are drawn from a set of linear functions of explained features, PGMExplainer is able to demonstrate the dependencies of explained features in form of conditional probabilities. Our theoretical analysis shows that the PGM generated by PGMExplainer includes the Markovblanket of the target prediction, i.e. including all its statistical information. We also show that the explanation returned by PGMExplainer contains the same set of independence statements in the perfect map. Our experiments on both synthetic and realworld datasets show that PGMExplainer achieves better performance than existing explainers in many benchmark tasks.
READ FULL TEXT
Comments
There are no comments yet.