Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Subgraph Concept Networks: Concept Levels in Graph Classification

About

The reasoning process of Graph Neural Networks is complex and considered opaque, limiting trust in their predictions. To alleviate this issue, prior work has proposed concept-based explanations, extracted from clusters in the model's node embeddings. However, a limitation of concept-based explanations is that they only explain the node embedding space and are obscured by pooling in graph classification. To mitigate this issue and provide a deeper level of understanding, we propose the Subgraph Concept Network. The Subgraph Concept Network is the first graph neural network architecture that distils subgraph and graph-level concepts. It achieves this by performing soft clustering on node concept embeddings to derive subgraph and graph-level concepts. Our results show that the Subgraph Concept Network allows to obtain competitive model accuracy, while discovering meaningful concepts at different levels of the network.

Lucie Charlotte Magister, Alexander Norcliffe, Iulia Duta, Pietro Lio• 2026

Related benchmarks

TaskDatasetResultRank
Graph ClassificationREDDIT BINARY
Accuracy91.23
144
Graph ClassificationMutagenicity
Accuracy80.53
35
Graph ClassificationGRID
Accuracy99.4
8
Graph ClassificationGrid-House
Model Accuracy89.5
8
Graph ClassificationHouse-Colour
Accuracy99.6
8
Graph ClassificationSTARS
Accuracy99.13
8
Showing 6 of 6 rows

Other info

Follow for update