Table 1 Impact of the weighting factor lambda between the cross-entropy classification loss and the smoothing mean-squared-error loss on the neural network, evaluation was done on the validation subset of the data.
Metric | λ = 0.15 | λ = 0.20 | λ = 0.25 | λ = 0.30 | λ = 0.35 | λ = 0.40 | λ = 0.50 |
|---|---|---|---|---|---|---|---|
Accuracy | 83.94 | 81.50 | 82.98 | 87.76 | 89.83 | 88.27 | 86.46 |
Edit | 77.56 | 76.03 | 73.67 | 80.54 | 86.81 | 82.12 | 78.65 |
ROC AUC | 97.37 | 96.64 | 97.15 | 98.60 | 98.82 | 98.53 | 98.33 |
PR AUC | 82.41 | 78.60 | 80.73 | 86.91 | 91.58 | 87.34 | 84.63 |