Table 9.

Comparison of the time and memory complexity of ConvGNN learning models.

Complexity	GCN (Singhal, 2020)	GraphSage (Hamilton et al., 2017)	CNN (Giulini and Potestio, 2019)
Time	$O\left(Kmd+Kn{d}^2\right)$	$O\left(r^{K}n{d}^2\right)$	N/A
Memory	$O\left(Knd+K{d}^2\right)$	$O\left(s{r}^{K}d+K{d}^2\right)$	N/A

Where n denotes the number of nodes in total. The total number of edges is denoted by m. The number of layers is denoted by K. The batch size is denoted by s. r is the number of neighbors that each node is sampled. For simplicity, the dimensions of the node's hidden features, represented by d, remain constant.