这是indexloc提供的服务,不要输入任何密码
Skip to main content
Log in

Table 4 Classification accuracy of multi-channel EEG signals under different sparsity constraints

From: Non-linear Feature Selection Based on Convolution Neural Networks with Sparse Regularization

Data

Backbone

Original accuracy

LSWFSNet-M(1)

LSWFSNet-M(2)

LSWFSNet-M(1+(2,1))m

DJC20131107

VGG-16

65.76

67.22 (9.35)

64.98 (9.03)

72.10 (11.61)

 

Alexnet

71.71

73.66 (7.74)

72.68 (10.00)

73.56 (10.32)

 

Googlenet

63.51

71.90 (18.71)

66.73 (22.58)

69.85 (22.90)

 

Resnet-34

40.39

42.24 (9.35)

42.83 (11.29)

45.07 (10.32)

 

Densenet-101

38.54

40.20 (16.13)

42.34 (11.94)

43.32 (14.52)

 

Efficientnet-B0

37.95

40.10 (10.65)

38.73 (10.00)

38.44 (10.65)

 

Mobilenet-V2

36.78

38.15 (11.94)

37.76 (12.58)

37.76 (14.52)

JL20140413

VGG-16

68.98

65.37 (11.61)

62.05 (11.29)

73.76 (5.81)

 

Alexnet

70.83

72.88 (11.29)

73.07 (7.74)

72.39 (8.06)

 

Googlenet

69.17

70.93 (12.58)

72.78 (21.61)

72.59 (23.23)

 

Resnet-34

48.88

51.22 (12.58)

44.00 (11.61)

52.00 (8.39)

 

Densenet-101

44.20

43.51 (14.84)

47.71 (7.74)

44.10 (13.23)

 

Efficientnet-B0

40.00

40.78 (12.26)

40.59 (11.61)

40.78 (21.61)

 

Mobilenet-V2

37.66

38.54 (13.87)

37.85 (8.71)

40.78 (12.58)

JL20140404

VGG-16

65.76

64.98 (8.39)

66.73 (10.65)

74.54 (11.94)

 

Alexnet

71.41

73.17 (9.35)

74.44 (9.35)

75.12 (10.32)

 

Googlenet

68.49

72.20 (17.10)

72.20 (24.19)

71.61 (21.29)

 

Resnet-34

39.02

44.49 (10.97)

45.95 (10.00)

46.24 (14.19)

 

Densenet-101

38.73

44.29 (11.29)

41.95 (21.94)

40.59 (8.71)

 

Efficientnet-B0

36.39

37.17 (10.65)

39.12 (10.65)

38.44 (10.32)

 

Mobilenet-V2

36.88

39.51 (11.61)

36.00 (20.00)

37.56 (13.87)

JJ20140629

VGG-16

86.73

85.17 (10.65)

84.88 (11.29)

87.32 (14.52)

 

Alexnet

85.46

87.32 (7.42)

83.61 (11.61)

84.59 (10.97)

 

Googlenet

83.71

85.46 (19.35)

83.22 (17.74)

83.51 (19.03)

 

Resnet-34

75.71

76.98 (17.42)

73.66 (10.65)

77.76 (23.87)

 

Densenet-101

67.32

69.56 (11.94)

64.68 (25.16)

68.29 (15.16)

 

Efficientnet-B0

57.27

58.93 (12.58)

58.73 (15.48)

59.02 (20.00)

 

Mobilenet-V2

41.66

45.27 (11.29)

42.05 (16.45)

45.37 (14.19)

LY2014018

VGG-16

79.81

83.90 (8.39)

80.00 (11.94)

88.68 (10.32)

 

Alexnet

90.63

91.61 (10.97)

92.20 (10.97)

93.07 (11.29)

 

Googlenet

86.73

88.68 (21.29)

88.00 (23.23)

88.49 (20.00)

 

Resnet-34

66.93

68.00 (17.42)

75.22 (15.81)

68.68 (11.61)

 

Densenet-101

56.29

58.54 (21.61)

60.49 (15.81)

64.49 (18.39)

 

Efficientnet-B0

47.61

50.05 (11.61)

49.56 (20.00)

52.49 (12.90)

 

Mobilenet-V2

38.34

41.17 (12.26)

46.83 (13.87)

40.88 (11.29)

LY20140506

VGG-16

61.56

61.57 (9.68)

65.85 (13.23)

58.93 (10.32)

 

Alexnet

82.63

85.07 (10.00)

86.05 (10.00)

79.02 (10.65)

 

Googlenet

73.07

75.61 (9.68)

78.34 (17.42)

78.73 (9.35)

 

Resnet-34

54.73

56.98 (11.29)

56.49 (12.58)

56.49 (9.35)

 

Densenet-101

55.02

55.71 (11.94)

55.61 (14.52)

54.15 (7.10)

 

Efficientnet-B0

43.02

46.54 (8.39)

43.02 (17.42)

47.32 (10.97)

 

Mobilenet-V2

37.66

39.02 (9.68)

39.90 (10.00)

42.15 (21.61)

MHW20131016

VGG-16

81.56

82.05 (8.06)

88.78 (8.71)

76.39 (10.65)

 

Alexnet

88.88

90.15 (9.68)

88.88 (10.00)

89.17 (6.45)

 

Googlenet

88.59

89.37 (21.61)

87.90 (13.87)

87.71 (15.16)

 

Resnet-34

74.83

77.66 (17.10)

79.51 (19.35)

77.37 (17.42)

 

Densenet-101

64.20

66.05 (9.03)

68.88 (10.97)

66.44 (14.84)

 

Efficientnet-B0

53.66

57.76 (12.26)

60.29 (13.87)

60.20 (19.35)

 

Mobilenet-V2

40.88

43.51 (9.68)

42.54 (8.71)

42.73 (12.58)

LQJ20140621

VGG-16

91.22

90.15 (10.00)

91.10 (10.65)

92.00 (10.97)

 

Alexnet

93.46

92.29 (13.23)

89.56 (11.61)

91.61 (9.35)

 

Googlenet

91.80

92.20 (18.71)

92.29 (21.29)

92.39 (19.03)

 

Resnet-34

83.51

86.05 (13.23)

80.00 (11.61)

87.22 (11.29)

 

Densenet-101

75.22

79.90 (18.06)

75.80 (7.42)

80.00 (19.68)

 

Efficientnet-B0

65.95

67.22 (15.48)

70.54 (9.68)

69.76 (16.45)

 

Mobilenet-V2

53.37

61.85 (10.65)

55.12 (8.71)

57.76 (13.55)

  1. The bold values indicates highest accuracy under same network, and the numbers in parentheses indicate the ratio of features screened out under different sparsity constraint