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

Table 3

Performance (5-fold CV AUC) comparison on full set of data.

Method	S1	S2	S3	S4	S5	S6	S7	S8	S9	S10	Mean  Std
PCA + LDA	0.8467	0.7596	0.8460	0.7678	0.8085	0.8324	0.8871	0.8733	0.7932	0.8353
xDAWN + LDA	0.8531	0.7504	0.8711	0.8198	0.8366	0.8017	0.8808	0.8957	0.7971	0.8380
DeepConvNet	0.9135	0.8603	0.9193	0.8456	0.9075	0.8634	0.9188	0.9071	0.8959	0.9167
EEGNet	0.9307	0.8568	0.9214	0.8589	0.8970	0.8597	0.8999	0.9272	0.9188	0.9262
MDRM	0.7901	0.6995	0.9170	0.7562	0.8858	0.8237	0.9377	0.8485	0.8996	0.9382
TSLDA	0.9298	0.8587	0.9059	0.6931	0.8619	0.8118	0.9464	0.8946	0.7898	0.9270
SPDNet	0.9453	0.8755	0.9307	0.8524	0.9366	0.8713	0.9480	0.9091	0.8999	0.9251
Ours	0.9521	0.8907	0.9519	0.8895	0.9400	0.9236	0.9428	0.9324	0.9451	0.9489