Epileptic Seizure Detection Using a Hybrid 1D CNN-Machine Learning Approach from EEG Data

Table 2

CNN hyperparameters for 3, 4, and 5-classes problems.

Bonn datasets					CHB-MIT datasets
Layers	2-class	3-class	4-class	5-class	2-class
Conv.	K = 4, Ks = 3, S = 1	K = 10, Ks  =  4, S = 1	K = 11, Ks  = 5, S = 1	K = 11, Ks  = 5, S  = 1	K = 11, Ks  = 3, S = 1
Max. Pooling	Ps = 4, S = 4	Ps  = 4, S = 4	Ps  = 4, S = 4	Ps  = 4, S = 4	Ps  = 4, S = 4
Conv.	—	K = 5, Ks  = 5, S = 1	K = 9, Ks  = 5, S = 1	K = 7, Ks  = 5, S  = 1	K  = 7, Ks  = 11, S  = 1
Max. Pooling	—	Ps  = 4, S = 4	Ps  = 4, S = 4	Ps  = 4, S  = 4	Ps  = 4, S  = 4
Flatten	F = 7720	F = 2405	F = 4329	F = 3367	F  = 1043
Dense	N  = 50	N  = 50	N  = 50	N  = 50	N  = 50
Dense	N  = 20	N  = 20	N  = 20	N  = 20	N  = 20
Dense	N  = 2	N  =  3	N  = 4	N  = 5	N  = 2