English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Graph neural networks (GNNs) are highly effective on a variety ofgraph-related tasks; however, they lack interpretability and transparency.Current explainability approaches are typically local and treat GNNs asblack-boxes. They do not look inside the model, inhibiting human trust in themodel and explanations. Motivated by the ability of neurons to detecthigh-level semantic concepts in vision models, we perform a novel analysis onthe behaviour of individual GNN neurons to answer questions about GNNinterpretability, and propose new metrics for evaluating the interpretabilityof GNN neurons. We propose a novel approach for producing global explanationsfor GNNs using neuron-level concepts to enable practitioners to have ahigh-level view of the model. Specifically, (i) to the best of our knowledge,this is the first work which shows that GNN neurons act as concept detectorsand have strong alignment with concepts formulated as logical compositions ofnode degree and neighbourhood properties; (ii) we quantitatively assess theimportance of detected concepts, and identify a trade-off between trainingduration and neuron-level interpretability; (iii) we demonstrate that ourglobal explainability approach has advantages over the current state-of-the-art-- we can disentangle the explanation into individual interpretable conceptsbacked by logical descriptions, which reduces potential for bias and improvesuser-friendliness.

Global Concept-Based Interpretability for Graph Neural Networks via Neuron Analysis