A new survey paper details the widespread applications of Graph Neural Networks (GNNs) across various domains, moving beyond their initial niche status. The paper organizes the field by deriving spectral and spatial formulations from shared principles and connecting their expressive power to the Weisfeiler-Leman hierarchy. It examines twelve application domains, including recommendation systems, drug discovery, and computer vision, analyzing graph construction choices, dominant architectures, and reported gains versus actual performance. The survey also highlights recurring patterns such as heterophily and scale impacting model performance across domains, the challenges of temporal graphs, and the gap between top-ranked architectures and those deployed in practice. AI
IMPACT Provides a comprehensive overview of GNN applications, aiding researchers and practitioners in understanding their capabilities and limitations across diverse fields.
RANK_REASON The cluster contains a survey paper on Graph Neural Networks.
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
