Table 5 Comparison of results (accuracies in %) obtained by the different optimization algorithms (OAs) for feature selection (FS) in our DSwarm-Net model and that obtained without any FS, by end-to-end CNN model.

From: A union of deep learning and swarm-based optimization for 3D human action recognition

OAs

UTD-MHAD

NTU RGB+D 60

HDM05

Cross subject

Cross view

Without FS

89.34

79.19

80.63

81.53

MVO56

96.26

84.67

89.16

90.41

PSO57

96.47

83.85

87.58

88.56

BAT58

95.74

84.01

88.52

87.42

GWO59

95.63

84.43

87.51

88.13

WOA60

95.26

84.22

88.13

86.87

FFA61

95.84

84.11

87.98

86.55

MFO62

95.53

83.89

89.51

89.93

ALO

98.13

85.45

89.98

90.67