Why does GCN use symmetric normalization on the adjacency matrix?

前端未结

关注

 0  902

In the GCN proposed by Thomas Kipf in 2017, symmetric normalization was used in propagation rule: $f\left(H^{(l)},&$