dbw
Supplementary Material On Provable Benefits of Depth in Training Graph Convolutional Networks
Finally in GCNs, we from discuss [26, conditions Before processing, where ov let er-smoothing first recall and the o notation ver-fitting we might defined happen in Section simultaneously 3. Recall .
SupplementaryMaterial OnProvableBenefitsofDepthinTraining GraphConvolutionalNetworks
Under the assumption thatγGCN 1, i.e., the graph is densely connected and the largest singular value ofweight matrices issmall, we havedM(H(`+1)) γGCN dM(H(`)).