Table 5.

Spelling accuracy with the models obtained 1 year before

Subject	Batch 1	Batch 2	Batch 3	Batch 4	60 Char test
S1	0.6	0.8	1.0	1.0	0.9
S2	0.6	0.8	0.8	1.0	0.95
S3	0.6	1.0	1.0	1.0	0.933
Average	0.60	0.8667	0.9333	1.0	0.928

The initial training set contains the EEG data of 8 character input collected 1 year before. For semi-supervised learning, each batch contains the unlabelled data with respect to the input of 5 characters. The final column gives the results on a 60-character independent test set