Skip to main content
. 2020 Oct 9;8:e10124. doi: 10.7717/peerj.10124

Table 2. Hyperparameters optimizations results.

Markers Upper limb type (three classes) Participants group (two classes)
Hyperparameters CNN metrices Hyperparameters CNN metrices
FN lr = 0.00017 lr = 0.0011
ep = 100 AUC = 85.54% ep = 224 AUC = 94.53%
fConv1 = 77 Accuracy = 69.23% fConv1 = 83 Accuracy = 88.5%
fConv2 = 10 Precision = 20% fConv2 = 100 Precision = 88.5%
dDrop1 = 0.25 Recall = 65.38% dDrop1 = 0.2484 Recall = 88.5%
outDense1 = 100 outDense1 = 30
LEP lr = 0.00037 lr = 0.00026
ep = 100 AUC = 89.53% ep = 216 AUC = 93.62%
fConv1 = 20 Accuracy = 80.77% fConv1 = 51 Accuracy = 92.6%
fConv2 = 10 Precision = 80.77% fConv2 = 10 Precision = 92.6%
dDrop1 = 0.1931 Recall = 80.77% dDrop1 = 0.0638 Recall = 92.6%
outDense1 = 72 outDense1 = 11
MPH lr = 0.00023 lr = 0.00004
ep = 100 AUC = 96.18% ep = 261 AUC = 99.5%
fConv1 = 100 Accuracy = 87.50% fConv1 = 87 Accuracy = 99%
fConv2 = 79 Precision = 90.91% fConv2 = 51 Precision = 98%
dDrop1 = 0.25 Recall = 83.33% dDrop1 = 0.1054 Recall = 98%
outDense1 = 44 outDense1 = 19
ACR lr = 0.01 lr = 0.0026
ep = 300 AUC = 83.98% ep = 202 AUC = 99.20%
fConv1 = 12 Accuracy = 58.33% fConv1 = 50 Accuracy = 96%
fConv2 = 54 Precision = 60.87% fConv2 = 100 Precision = 96%
dDrop1 = 0.004 Recall = 58.33% dDrop1 = 0.2435 Recall = 96%
outDense1 = 10 outDense1 = 26
LEP, MPH lr = 0.00006 lr = 0.0064
ep = 300 AUC = 94.62% ep = 300 AUC = 99.5%
fConv1 = 75 Accuracy = 87.5% fConv1 = 100 Accuracy = 99%
fConv2 = 26 Precision = 89.47% fConv2 = 10 Precision = 98%
dDrop1 = 0 Recall = 70.83% dDrop1 = 0.1797 Recall = 98%
outDense1 = 98 outDense1 = 10
FN, LEP, MPH lr = 0.00012 lr = 0.00003
ep = 300 AUC = 90% ep = 300 AUC = 98.4%
fConv1 = 100 Accuracy = 72% fConv1 = 95 Accuracy = 92%
fConv2 = 10 Precision = 75% fConv2 = 10 Precision = 92%
dDrop1 = 0 Recall = 72% dDrop1 = 0.25 Recall = 92%
outDense1 = 10 outDense1 = 100
LEP, MPH, ACR lr = 0.00005 lr = 0.00017
ep = 296 AUC = 83.11% ep = 164 AUC = 89.41%
fConv1 = 62 Accuracy = 68% fConv1 = 100 Accuracy = 86.9%
fConv2 = 87 Precision = 68.11% fConv2 = 78 Precision = 86.9%
dDrop1 = 0.0845 Recall = 68.11% dDrop1 = 0.0126 Recall = 86.9%
outDense1 = 72 outDense1 = 85
FN, LEP, MPH, ACR lr = 0.00019 lr = 0.000074
ep = 300 AUC = 96.33% ep = 290 AUC = 99.7%
fConv1 = 10 Accuracy = 77.27% fConv1 = 22 Accuracy = 99%
fConv2 = 65 Precision = 77.27% fConv2 = 39 Precision = 98.5%
dDrop1 = 0.25 Recall = 77.27% dDrop1 = 0.2491 Recall = 98.5%
outDense1 = 100 outDense1 = 99