Table 3 Classification accuracy of single-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-S(1) | LSWFSNet-S(2) | LSWFSNet-S(1+2) |
|---|---|---|---|---|---|
JL20140404 | VGG-16 | 77.27 | 82.34 (16.13) | 78.05 (17.74) | 77.95 (8.06) |
Alexnet | 69.07 | 71.80 (9.68) | 70.93 (14.52) | 76.98 (8.06) | |
Googlenet | 57.07 | 65.56 (20.97) | 65.37 (14.52) | 60.68 (24.19) | |
Resnet-34 | 35.22 | 40.49 (8.06) | 36.20 (3.23) | 45.56 (14.52) | |
Densenet-101 | 34.93 | 37.27 (16.13) | 38.44 (12.90) | 39.12 (9.68) | |
Efficientnet-B0 | 37.76 | 40.10 (20.97) | 37.56 (14.52) | 39.41 (14.52) | |
Mobilenet-V2 | 36.68 | 37.27 (24.19) | 35.80 (14.52) | 37.95 (17.74) | |
JL20140419 | VGG-16 | 78.54 | 78.54 (6.45) | 74.34 (8.06) | 76.39 (11.29) |
Alexnet | 70.83 | 72.49 (9.68) | 74.05 (4.84) | 70.93 (4.84) | |
Googlenet | 65.76 | 68.98 (22.58) | 68.49 (6.45) | 68.68 (25.81) | |
Resnet-34 | 51.61 | 54.54 (25.81) | 51.51 (9.68) | 52.39 (11.29) | |
Densenet-101 | 42.63 | 45.95 (8.06) | 44.98 (11.29) | 48.88 (19.35) | |
Efficientnet-B0 | 43.61 | 47.71 (11.29) | 44.59 (16.13) | 44.88 (19.35) | |
Mobilenet-V2 | 37.37 | 39.12 (6.45) | 37.07 (12.90) | 39.41 (9.68) | |
JJ20140603 | VGG-16 | 86.73 | 89.27 (12.90) | 84.49 (6.45) | 85.95 (8.06) |
Alexnet | 80.78 | 83.90 (12.90) | 85.46 (4.84) | 84.39 (9.68) | |
Googlenet | 68.98 | 76.20 (19.35) | 81.95 (6.45) | 84.49 (14.52) | |
Resnet-34 | 45.76 | 52.98 (12.90) | 50.73 (4.84) | 49.95 (12.90) | |
Densenet-101 | 41.76 | 43.32 (27.42) | 40.88 (6.45) | 44.49 (11.29) | |
Efficientnet-B0 | 41.66 | 43.12 (16.13) | 42.05 (22.58) | 44.49 (9.68) | |
Mobilenet-V2 | 41.56 | 43.80 (9.68) | 42.73 (9.68) | 42.93 (29.03) | |
JJ20140611 | VGG-16 | 90.15 | 91.90 (6.45) | 92.59 (3.23) | 93.37 (4.84) |
Alexnet | 86.63 | 88.68 (9.68) | 81.95 (9.68) | 88.49 (17.74) | |
Googlenet | 82.44 | 84.49 (17.74) | 78.24 (12.90) | 86.54 (27.42) | |
Resnet-34 | 66.73 | 70.44 (19.35) | 73.76 (4.84) | 79.80 (19.35) | |
Densenet-101 | 48.88 | 53.95 (16.13) | 53.46 (3.23) | 49.85 (9.68) | |
Efficientnet-B0 | 41.27 | 46.05 (19.35) | 44.88 (19.35) | 45.37 (16.13) | |
Mobilenet-V2 | 36.78 | 39.80 (4.84) | 39.90 (12.90) | 38.83 (11.29) | |
LY20140506 | VGG-16 | 82.73 | 88.88 (6.45) | 83.51 (4.84) | 87.41 (6.45) |
Alexnet | 81.27 | 86.73 (8.06) | 81.66 (8.06) | 86.63 (9.68) | |
Googlenet | 73.85 | 78.05 (25.81) | 78.83 (17.74) | 79.02 (24.19) | |
Resnet-34 | 64.59 | 67.02 (16.13) | 65.46 (8.06) | 67.32 (25.81) | |
Densenet-101 | 55.61 | 59.51 (11.29) | 60.00 (9.68) | 57.66 (25.81) | |
Efficientnet-B0 | 44.20 | 49.85 (14.52) | 51.41 (16.13) | 50.05 (17.74) | |
Mobilenet-V2 | 41.85 | 44.88 (12.90) | 39.51 (12.90) | 47.02 (11.29) | |
LY20140411 | VGG-16 | 74.73 | 75.02 (14.52) | 63.90 (19.35) | 84.59 (8.06) |
Alexnet | 82.93 | 84.00 (16.13) | 84.00 (17.74) | 85.66 (17.74) | |
Googlenet | 72.68 | 78.93 (30.65) | 79.80 (8.06) | 81.85 (22.58) | |
Resnet-34 | 48.59 | 52.59 (11.29) | 53.07 (11.29) | 51.32 (20.97) | |
Densenet-101 | 44.39 | 54.54 (9.68) | 46.15 (3.23) | 48.00 (14.52) | |
Efficientnet-B0 | 46.93 | 53.37 (30.65) | 52.78 (20.97) | 50.44 (19.35) | |
Mobilenet-V2 | 40.39 | 42.24 (11.29) | 40.59 (6.45) | 43.71 (25.81) | |
MHW20130712 | VGG-16 | 71.12 | 76.00 (6.45) | 65.46 (12.90) | 90.44 (12.90) |
Alexnet | 78.93 | 87.51 (6.45) | 80.98 (1.61) | 83.61 (17.74) | |
Googlenet | 68.29 | 77.76 (24.19) | 86.15 (6.45) | 81.66 (33.87) | |
Resnet-34 | 50.15 | 59.61 (24.19) | 50.05 (1.29) | 60.00 (9.68) | |
Densenet-101 | 42.73 | 52.29 (24.19) | 42.24 (3.23) | 44.68 (4.84) | |
Efficientnet-B0 | 41.27 | 58.63 (14.52) | 41.07 (4.84) | 44.39 (20.97) | |
Mobilenet-V2 | 40.88 | 43.12 (17.74) | 40.59 (11.29) | 41.85 (12.90) |