Table 4.

Decoding accuracies residual networks and difference to deep ConvNets

Dataset	Frequency range [Hz]	Accuracy	Difference to deep	P value
BCIC IV 2a	0–38	67.7	−3.2	0.13
BCIC IV 2a	4–38	60.8	−9.3	0.004**
HGD	0–125	88.9	−3.5	0.020*
HGD	4–125	89.8	−1.6	0.54
Combined	0–fend	80.6	−3.4	0.004**
Combined	4–fend	78.5	−4.9	0.01*

BCIC IV 2a: BCI competition IV dataset 2a. HGD: High‐Gamma Dataset. Accuracy is mean accuracy in percentage. P value from Wilcoxon signed‐rank test for the statistical significance of the differences to the deep ConvNet (cropped training). Accuracies were always slightly worse than for the deep ConvNet, statistically significantly different for both frequency ranges on the combined dataset.