New framework enables exact Aumann-Shapley attribution for GNNs

Researchers have developed APEX, a novel framework designed to provide exact Aumann-Shapley attributions for graph neural networks (GNNs). This framework utilizes a specialized GNN architecture called PolyGIN, which maintains a polynomial form for model scores. By leveraging this polynomial structure, APEX enables exact computation of attribution integrals using Gauss--Legendre quadrature, significantly reducing the number of evaluations needed compared to traditional approximation methods. Experiments demonstrate that PolyGIN achieves competitive predictive performance while the APEX framework offers higher attribution fidelity. AI

IMPACT Enhances explainability for graph neural networks, potentially improving trust and debugging in AI applications that use graph data.

RANK_REASON Academic paper detailing a new framework and model architecture for GNN explainability. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• APEX
• arXiv
• Aumann--Shapley
• Gauss--Legendre method
• graph neural networks
• Integrated Gradients

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →