Toward Measuring Target Perception: First-Order and Second-Order Deep Network Pipeline for Classification of Fixation-Related Potentials

Table 4

Performance (5-fold CV AUC) difference between the midlevel SPD matrices, respectively, aggregated by outputs from the first and the second convolutional blocks of EEGNet.

	S1	S2	S3	S4	S5	S6	S7	S8	S9	S10	MeanStd
Second block-first block (ours)	−0.0306	−0.0032	0.0095	−0.0195	−0.0209	−0.0490	0.0070	−0.0088	−0.0163	−0.0002	−0.01320.0180