Table 2 Comparing our full method to methods which omit one technique: ten runs best validation
From: Learning to detect the onset of slow activity after a generalized tonic–clonic seizure
Method variant | ROC–AUC | Sensitivity | Specificity | Accuracy |
|---|---|---|---|---|
Baseline | 0.639 | 0.613 | 0.562 | 0.575 |
W/o batch z-score | 0.667 | 0.202 | 0.886 | 0.708 |
W/o \(L_2\) | 0.680 | 0.539 | 0.794 | 0.727 |
Batch size = 64 | 0.696 | 0.488 | 0.759 | 0.689 |
Batch size = 32 | 0.725 | 0.636 | 0.667 | 0.659 |
W/o anti-aliasing | 0.728 | 0.505 | 0.774 | 0.704 |
W/o discount | 0.749 | 0.464 | 0.832 | 0.736 |
Full method | 0.768 | 0.486 | 0.884 | 0.781 |