New GDCE-I method enhances explainability for Graph Neural Networks

Researchers have developed a new method called Graph Diffusion Counterfactual Explanation via Inversion (GDCE-I) to generate counterfactual explanations for Graph Neural Networks (GNNs). This approach uses a discrete denoising diffusion model with a novel inversion scheme to produce edits that are both on the data manifold and explore the full edit space. GDCE-I aims to provide faithful, sufficient, and understandable explanations, addressing limitations in existing methods that compromise on data distribution adherence or search completeness. Evaluations across four benchmarks demonstrate GDCE-I's superior performance, particularly in the molecular domain where it produces interpretable, in-distribution solutions. AI

IMPACT Enhances the interpretability and trustworthiness of GNNs in critical applications.

RANK_REASON The cluster contains a research paper detailing a new method for explaining AI models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• arXiv
• discrete diffusion inversion
• GDCE-I
• Graph Diffusion Counterfactual Explanation via Inversion
• Graph Neural Networks

• paper
• model release

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