Hybrid Low-Order and Higher-Order Graph Convolutional Networks

Table 7

Citation network classification test accuracy.

Approaches	Cora	CiteSeer	PubMed
ManiReg [31]	59.5	60.1	70.7
SemiEmb [32]	59.0	59.6	71.1
LP [33]	68.0	45.3	63.0
DeepWalk [34]	67.2	43.2	65.3
ICA [35]	75.1	69.1	73.9
Planetoid [30]	75.7	64.7	77.2
GCN [6]	81.5	70.3	79.0
HO-3 [14]	81.6 ± 0.47	71.2 ± 0.94	80.0 ± 0.64
HO-4 [14]	81.6 ± 0.63	71.2 ± 0.84	80.1 ± 0.65
MixHop [15]	81.8 ± 0.62	71.4 ± 0.81	80.0 ± 1.10
MixHop (learned) [15]	81.9 ± 0.40	71.4 ± 0.81	80.8 ± 0.58
HLHG-2 (ours)	82.7 ± 0.28	71.5 ± 0.22	79.1 ± 0.18
HLHG-3 (ours)	82.7 ± 0.29	71.5 ± 0.39	79.3 ± 0.15