. 2023 Mar 10;18(3):e0279604. doi: 10.1371/journal.pone.0279604

Table 1. Different graph neural network models based on MPNN framework.

Model	Message Passing Function	Message Update Function
GCN [2]	${m^{(t + 1)} = \tilde{D}}^{- 1 / 2} \tilde{A} {\tilde{D}}^{- 1 / 2} H^{(t)} W^{(t)}$	$H^{(t + 1)} = R e L U (m^{(t + 1)})$
GraphSAGE-mean [25]	$m^{(t + 1)} = W^{(t)} \cdot M E A N ({h_{i}^{(t)}} \cup {h_{j}^{(t)}, v_{j} \in N (v_{i})})$	$h_{i}^{(t + 1)} = R e L U (m^{(t + 1)})$
R-GCN [26]	$m^{(t + 1)} = \sum_{r \in R} \sum_{j \in N_{i}^{(r)}} \frac{1}{c_{i, r}} W_{r}^{(t)} h_{j}^{(t)} + W_{0}^{(t)} h_{i}^{(t)}$	$h_{i}^{(t + 1)} = R e L U (m^{(t + 1)})$
DGCNN [21]	${m^{(t + 1)} = \tilde{D}}^{- 1} \tilde{A} H^{(t)} W^{(t)}$	H^(t+1) = tanh(m^(t+1))
DiffPool [27]	${m^{(t + 1)} = \tilde{D}}^{- 1 / 2} \tilde{A} {\tilde{D}}^{- 1 / 2} H^{(t)} W^{(t)}$	H^(t+1) = ReLU(m^(t+1))
GIN [17]	$m^{(t + 1)} = (1 + ϵ^{(t + 1)}) \cdot h_{i}^{(t)} + \sum_{v_{j} \in N (v_{i})} h_{j}^{(t)}$	$h_{i}^{(t + 1)} = {M L P}^{(t + 1)} (m^{(t + 1)})$