Graph Neural Network and Its Relation with WL Technique

  • Unique Paper ID: 172411
  • Volume: 11
  • Issue: 8
  • PageNo: 3029-3039
  • Abstract:
  • The research revolved around the theoretical rationale and investigation of the WL algorithm and graph neural network. The study's two main objectives are to analyze the expressive power of graph neural networks and explore if they may outperform higher order WL algorithms. To effectively achieve the paper's goal, the study used non-experimental approach layered on top of a quantitative methodology. Along with searching for all the notable variants that are currently garnering attention from the perspective of theoretical analysis, the research also looks into the use of higher-order GNN and WL tests. Examining GNN's strength in conjunction with WL tests was the first step in the study process. From there, greater-order GNN and k-dimensional WL tests were investigated. Furthermore, the study emphasizes the challenges and limitations associated with conventional GNN and WL approaches, which may be addressed by applying more advanced GNN and WL techniques. Furthermore, when feature embedding is included, the expressive capability of GNNs outperforms that of the 1-WL test. However, more research is advised to fully examine the expressive capability of GNN and its higher variations.

Cite This Article

  • ISSN: 2349-6002
  • Volume: 11
  • Issue: 8
  • PageNo: 3029-3039

Graph Neural Network and Its Relation with WL Technique

Related Articles