Skip to main content

Table 5 Classification accuracies in percent and standard deviations (Number of iterations for HGK-WL and HGK-SP: 20 (100 for SYNTHIE), OOM— Out of Memory

From: A survey on graph kernels

Kernel

Dataset

Average

 

ENZYMES

FRANKENSTEIN

PROTEINS

SyntheticNew

Synthie

 

SP+RBF

71.0 ±0.8

72.8 ±0.2

76.6 ±0.5

96.2 ±0.4

52.8 ±1.8

73.9

HGK-SP

71.3 ±0.9

70.1 ±0.3

77.5 ±0.4

96.5 ±0.6

94.3 ±0.5

81.9

HGK-WL

67.6 ±1.0

73.6 ±0.4

76.7 ±0.4

98.8 ±0.3

96.8 ±0.5

82.7

GH

68.8 ±1.0

68.5 ±0.3

72.3 ±0.3

85.1 ±1.0

73.2 ±0.8

73.6

GI

71.7 ±0.8

76.3 ±0.3

76.9 ±0.5

83.1 ±1.1

95.8 ±0.5

80.8

P2K

69.2 ±0.4

OOM

73.5 ±0.5

91.7 ±0.9

50.2 ±1.9

71.2

  1. The highest accuracy value is highlighted in bold for each dataset