Table 1 Comparing our full method to methods which omit one technique: ten runs \( \mu \pm \sigma \)
From: Learning to detect the onset of slow activity after a generalized tonic–clonic seizure
Method varriant | ROC–AUC | Sensitivity | Specificity | Accuracy |
|---|---|---|---|---|
Baseline | \(0.600 \pm 0.024\) | \(0.446 \pm 0.081 \) | \(0.689 \pm 0.057\) | \(0.626 \pm 0.026\) |
Batch size = 64 | \(0.646 \pm 0.027\) | \(0.468 \pm 0.064 \) | \(0.732 \pm 0.042\) | \(0.664 \pm 0.022\) |
W/o \(L_2\) | \(0.651 \pm 0.018\) | \(0.486 \pm 0.064\) | \(0.747 \pm 0.049\) | \(0.679 \pm 0.028\) |
W/o batch z-score | \(0.658 \pm 0.034\) | \(0.379 \pm 0.060 \) | \(0.801 \pm 0.055\) | \(0.691 \pm 0.034\) |
Batch size = 32 | \(0.689 \pm 0.032\) | \(0.500 \pm 0.069 \) | \(0.759 \pm 0.048\) | \(0.692 \pm 0.027\) |
W/o anti-aliasing | \(0.708 \pm 0.017\) | \(0.527 \pm 0.066 \) | \(0.767 \pm 0.051\) | \(0.708 \pm 0.024\) |
W/o discount | \(0.712 \pm 0.026 \) | \(0.462 \pm 0.032 \) | \(0.825 \pm 0.037 \) | \(0.731 \pm 0.023\) |
Full method | \({0.725 \pm 0.025 }\) | \({0.448 \pm 0.063}\) | \({0.850 \pm 0.032}\) | \({0.746 \pm 0.016}\) |