Table 2.

Description of dataset splitting.

	Number of snippets			Participant IDs
	FM+	FM−	Total
Total	956 (53.6%)	828 (46.4%)	1784
Validation	237	202	439 (24.6%)	7, 10, 14, 15, 17, 18, 34, 44, 47, 50, 51
Set 1
Training	471	414	885 (49.6%)	1, 5, 8, 9, 12, 16, 19, 20, 21, 22, 23, 27, 28, 29, 30, 33, 36, 37, 38, 39, 40, 42, 43, 45, 48, 49
Testing	248	212	460 (25.8%)	3, 4, 26, 31, 32, 35, 41, 46
Set 2
Training	482	412	905 (50.7%)	1, 4, 8, 16, 19, 21, 23, 26, 28, 29, 31, 32, 33, 36, 39, 42, 45, 46, 49
Testing	237	203	440 (24.7%)	3, 5, 9, 12, 20, 22, 27, 30, 35, 37, 38, 40, 41, 43, 48
Set 3
Training	487	425	912 (51.1%)	1, 3, 4, 5, 9, 16, 19, 20, 23, 26, 27, 29, 31, 32, 38, 39, 40, 42, 49
Testing	232	201	433 (24.3%)	8, 12, 21, 22, 28, 30, 33, 35, 36, 37, 41, 43, 45, 46, 48
Set 4
Training	483	422	905 (50.7%)	1, 3, 4, 8, 9, 16, 19, 20, 23, 26, 28, 31, 32, 35, 37, 38, 40, 41, 43, 46
Testing	236	204	440 (24.7%)	5, 12, 21, 22, 27, 29, 30, 33, 36, 39, 42, 45, 48, 49
Set 5
Training	486	421	907 (50.8%)	1, 4, 5, 9, 19, 20, 21, 23, 26, 27, 29, 31, 35, 36, 38, 40, 41, 42, 45, 46, 48
Testing	233	205	438 (24.6%)	3, 8, 12, 16, 22, 28, 30, 32, 33, 37, 39, 43, 49

The validation set was used for hyperparameter tuning whereas sets 1–5 were used for cross-validation and evaluation of network architectures.