Graph Neural Networks suffer from severe scalability issue, which prevents them from being adapted to large-scale graphs.
Node-wise Sampling based Methods
Layer-wise Sampling based Methods
Subgraph-wise Sampling based Methods
Conclusion
Further Reading