Table 2. The main features of different GNN models.
GNN models | Main features |
---|---|
GNN (Scarselli et al., 2009) | Fixed-point iteration method |
The same parameters are used in feature aggregation | |
GCN (Bruna et al., 2013) | Aggregation based on degree matrix and adjacency matrix |
More advanced operations to extract node information | |
GraphSAGE (Hamilton, Ying & Leskovec, 2017) | Mini batch training |
Three aggregator, mean, LSTM and pooling | |
GAT (Veličković et al., 2017) | Adding attention mechanism of transformer |
More interpretable information |