Skip to main content

Table 1 Classification accuracies for labeled graph datasets

From: Sequence-to-sequence modeling for graph representation learning

Datasets

MUTAG

PTC

Enzymes

fig1

Nci1

Nci109

 

# Graphs

188

344

600

1113

4110

4127

 

EntropyWL1

2.72

4.03

5.45

5.54

4.21

4.22

 

EntropyWL2

4.65

7.09

12.60

13.26

8.23

8.25

 

ClusteringCoef

0

0.0025

0.453

0.51

0.003

0.003

 

Avg. Nodes

17.9

25.5

32.6

39.1

29.8

29.6

 

# Labels

7

19

2

3

37

38

 

# Classes

2

2

6

2

2

2

Avg Rank

SPK (BK05)

85.2 ±2.4

58.2 ± 2.4

40.1 ± 1.5

75.1 ± 0.5

73.0 ± 0.2

73.0 ± 0.2

7.5

RWK (G+03)

83.7 ±1.5

57.8 ±1.3

24.2 ±1.6

74.2 ±0.4

—

—

—

GK (S+09)

81.7 ± 2.1

57.2 ± 1.4

26.6 ±0.9

71.7 ±0.5

62.3 ± 0.2

62.6 ± 0.1

8.5

WLSK (S+11)

80.7 ±3.0

57.0 ±2.0

53.1 ±1.1

72.9 ±0.5

80.1 ±0.5

80.2 ±0.3

8.5

node2vec (G+16)

82.01 ±1.0

55.60 ±1.4

19.42 ±2.3

70.76 ±1.2

61.91 ±0.3

61.53 ±0.9

9.5

DGK (YW15)

87.4 ±2.7

60.1 ±2.5

53.4 ±0.9

75.7 ±0.5

80.3 ±0.4

80.3 ±0.3

5.6

PSCN (N+16)

9 2 . 6

62.3

—

75.9

78.6

—

—

WL-OA (K+16)

86.0 ±1.7

63.6 ±1.5

59.9 ±1.1

76.4 ±0.4

86.1 ±0.2

86.3 ±0.2

3.3

GCN (KI+17)

86.3 ±2.1

63.28 ±3.9

56.6 ±3.5

75.9 ±2.8

81.1 ±1.6

80.7 ±1.8

5.0

graph2vec (N+17)

83.15 ±9.2

60.17 ±6.9

—

73.30 ±2.0

73.22 ±1.9

74.26 ±1.5

—

WL PM (NI+17)

87.77 ±0.8

61.41 ±0.8

55.55 ±0.5

—

86.40 ±0.2

85.34 ±0.2

—

LWL (M+17)

85.2 ±1.6

64.7 ±0.2

61.8 ±1.2

76.4 ±0.7

83.1 ±0.2

82.0 ±0.3

3.0

DGCNN (Z+18)

85.83 ±1.6

58.59 ±2.4

—

75.54 ±0.9

74.44 ±0.4

—

—

SGR (T+18)

86.97

—

33.67

73.83

—

—

—

S2S-N2N-PP

89.86 ±1.1

64.54 ±1.1

6 3 . 9 6 ± 0 . 6

76.51 ±0.5

83.72 ±0.4

83.64 ±0.3

2 . 5

supervised

89.91 ±1.5

6 5 . 7 2 ± 1 . 2

57.48 ±0.8

7 7 . 2 8 ± 0 . 9

8 6 . 6 5 ± 0 . 6

8 6 . 6 1 ± 0 . 5

1 . 5