Table 5 Experimental results for Session 1(S1S + S1R) and Session 2(S2s + S2R) based on different pre-trained models
From: An evaluation of transfer learning models in EEG-based authentication
Model | Comparison of performance metrics (Averaged% ± Standard deviation) | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
Session1: S1S(training) + S1R(testing) | Session 2: S2s(training) + S2R(testing) | |||||||||
Acc. | Pre. | Sens. | Spec. | F1 | Acc. | Pre. | Sens. | Spec. | F1 | |
GoogLeNet | 49.84 ± 1.40 | 48.30 ± 3.39 | 49.76 ± 1.61 | 98.21 ± 0.05 | 49.67 ± 9.73 | 41.45 ± 0.52 | 39.49 ± 0.33 | 41.19 ± 0.50 | 97.91 ± 0.02 | 37.90 ± 0.03 |
Inception-V3 | 54.07 ± 0.20 | 50.87 ± 0.63 | 54.00 ± 0.35 | 98.36 ± 0.01 | 48.46 ± 0.76 | 48.11 ± 0.18 | 42.87 ± 0.83 | 48.08 ± 0.46 | 98.15 ± 0.01 | 42.80 ± 0.33 |
ResNet-50 | 55.59 ± 1.45 | 51.20 ± 2.67 | 55.72 ± 1.49 | 98.42 ± 0.05 | 49.90 ± 2.23 | 48.06 ± 0.02 | 42.98 ± 2.47 | 48.08 ± 0.09 | 98.15 ± 0.01 | 42.69 ± 0.41 |
ResNet-101 | 58.48 ± 1.01 | 51.95 ± 2.09 | 58.74 ± 1.11 | 98.52 ± 0.04 | 52.28 ± 0.85 | 51.19 ± 0.67 | 43.47 ± 0.35 | 51.23 ± 0.51 | 98.26 ± 0.02 | 45.43 ± 0.31 |
Efficient-B0 | 50.40 ± 0.43 | 47.37 ± 2.18 | 50.15 ± 0.41 | 98.23 ± 0.02 | 45.58 ± 0.73 | 43.91 ± 1.08 | 42.04 ± 0.58 | 44.14 ± 1.04 | 98.00 ± 0.04 | 40.39 ± 1.36 |
DenseNet-201 | 55.30 ± 1.59 | 53.21 ± 2.40 | 55.51 ± 1.50 | 98.37 ± 0.11 | 49.84 ± 1.08 | 50.38 ± 0.74 | 45.88 ± 4.55 | 50.40 ± 0.79 | 98.23 ± 0.03 | 45.11 ± 1.76 |