A Multichannel 2D Convolutional Neural Network Model for Task-Evoked fMRI Data Classification

Table 3

A comparison of model units, parameters, and training time.

Model	Unit number input to fully connected layer	Total number of parameters in models	Training time (S) (mean ± std)	Total number of epochs (mean ± std)
mv2D CNN	16,800	2,223,877	909 ± 134	54 ± 8
3D CNN	352,800	45,174,181	1156 ± 185	39 ± 6
3D SepConv	352,800	45,161,301	1601 ± 196	41 ± 5
1D CNN	79,296	10,474,501	834 ± 157	39 ± 7
s2D CNN	16,800	2,236,837	565 ± 102	31 ± 6
M2D CNN	47,712	6,355,717	1074 ± 348	39 ± 13