New paper details over-squashing problem in Graph Neural Networks

This paper provides a comprehensive overview of the over-squashing problem in Graph Neural Networks (GNNs), a challenge that limits accuracy when long-range dependencies between graph nodes are required. The authors categorize existing approaches to mitigate this issue and discuss its relationship with expressive power and over-smoothing. The work also outlines methods for verifying the effectiveness of these mitigation techniques and identifies open questions for future research. AI

IMPACT This research clarifies a key limitation in Graph Neural Networks, potentially guiding future development of more capable models for graph-based learning tasks.

RANK_REASON The item is an academic paper detailing a specific problem within Graph Neural Networks. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• DagsHub
• Ethan A Shirley
• Graph Neural Networks
• Hugging Face
• IArxiv
• MPNNs
• OpenStreetMap
• over-squashing
• Royal Institute of Technology

• paper
• other

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