Skip to main content
. Author manuscript; available in PMC: 2023 Jun 1.
Published in final edited form as: Adv Neural Inf Process Syst. 2022 Dec;35:1909–1922.

Table 7:

Results on the test data sets with regard to robustness. Bold values indicate the best result. 10% of all vertexes are used for training.

Cora Citeseer ModelNet40
Random Net Minmax Random Net Minmax Random Net Minmax
SetGNN 66.87 ± 1.33 66.26 ± 1.54 66.58 ± 1.02 62.89 ± 1.57 62.81 ± 1.32 62.21 ± 1.64 95.74 ± 0.22 95.41 ± 0.28 93.33 ± 0.26
A2 71.90 ± 1.63 71.16 ± 0.92 70.86 ± 1.22 66.41 ± 1.08 65.38 ± 1.47 64.69 ± 0.98 96.09 ± 0.17 95.52 ± 0.24 93.64 ± 0.26
A4 72.11 ± 1.60 70.49 ± 1.29 70.52 ± 1.39 65.94 ± 1.24 65.15 ± 1.70 64.12 ± 1.19 95.79 ± 0.27 95.44 ± 0.25 93.35 ± 0.24
A6 72.15 ± 1.70 71.94 ± 1.48 71.98 ± 1.36 66.60 ± 1.61 65.68 ± 1.09 65.51 ± 1.13 96.58 ± 0.24 96.23 ± 0.23 94.82 ± 0.33
NTU2012 House (0.6) House (1.0)
Random Net Minmax Random Net Minmax Random Net Minmax
SetGNN 73.84 ± 2.18 73.38 ± 1.36 70.71 ± 1.89 67.16 ± 2.55 68.88 ± 2.68 64.78 ± 2.20 56.86 ± 1.93 59.95 ± 1.92 56.52 ± 2.52
A2 74.50 ± 2.03 73.86 ± 1.84 71.40 ± 1.64 67.71 ± 2.94 69.59 ± 2.32 65.23 ± 2.89 57.74 ± 2.70 60.73 ± 2.30 57.00 ± 1.94
A4 73.73 ± 1.59 73.72 ± 1.59 71.06 ± 1.53 67.55 ± 2.41 68.85 ± 1.38 64.97 ± 3.35 57.47 ± 2.72 60.10 ± 1.74 56.65 ± 2.26
A6 75.06 ± 1.97 74.37 ± 1.99 72.09 ± 1.98 69.88 ± 3.27 73.14 ± 2.71 68.84 ± 2.71 60.06 ± 2.07 62.41 ± 1.77 58.76 ± 2.24