Table 4:

Testing accuracies (%) for graph classification. We highlight the result of the best performed model for each dataset. Our GAT-GC (f-Scaled) model achieves the top 2 on all 6 datasets.

	Datasets	MUTAG	PROTEINS	ENZYMES	NCI1	RE-B	RE-M5K
Baselines	WL	82.05 ± 0.36	74.68 ± 0.49	52.22 ± 1.26	82.19 ±0.18	81.10 ± 1.90	49.44 ± 2.36
	PSCN	88.95 ± 4.37	75.00 ±2.51	-	76.34 ± 1.68	86.30 ±1.58	49.10 ±0.70
	DGCNN	85.83 ± 1.66	75.54 ± 0.94	51.00 ±7.29	74.44 ± 0.47	76.02 ± 1.73	48.70 ± 4.54
	GIN	89.40 ± 5.60	76.20 ± 2.80	-	82.70 ± 1.70	92.40 ± 2.50	57.50 ± 1.50
	CapsGNN	86.67 ± 6.88	76.28 ± 3.63	54.67 ± 5.67	78.35 ± 1.55	-	52.88 ± 1.48
	GAT-GC (f-Scaled)	90.44 ± 6.44	76.81 ± 3.77	58.45 ± 6.35	82.28 ± 1.81	92.57 ± 2.06	57.22 ± 2.20