Table 6.

The CNN architectures used in the experiment.

Layer		CNN1	CNN2	CNN3	CNN4	Alexnet	VGGNet-9
Input		1 × 2071	1 × 2071	1 × 2071	1 × 2071	1 × 2071	1 × 2071
Convolution	1	32 kernels in size 1 × 3 with max pooling	32 kernels in size 1 × 3 with max pooling	32 kernels in size 1 × 3 with max pooling	32 kernels in size 1 × 3 with max pooling	96 kernels in size 1 × 11 with max pooling	64 kernels in size 1 × 5
	2	—	32 kernels in size 1 × 3 with max pooling	32 kernels in size 1 × 3 with max pooling	32 kernels in size 1 × 3 with max pooling	256 kernels in size 1 × 5 with max pooling	64 kernels in size 1 × 3 with max pooling
	3	—	—	32 kernels in size 1 × 3 with max pooling	32 kernels in size 1 × 3 with max pooling	384 kernels in size 1 × 3 with max pooling	128 kernels in size 1× 3
	4	—	—	—	32 kernels in size 1 × 3 with max pooling	384 kernels in size 1 × 3	128 kernels in size 1 × 3 with max pooling
	5	—	—	—	—	256 kernels in size 1 × 3	256 kernels in size 1 × 3
	6	—	—	—	—	—	256 kernels in size 1 × 3
	7	—	—	—	—	—	256 kernels in size 1 × 3 with max pooling
Fully connected	1	512 nodes, ReLu	512 nodes, ReLu	512 nodes, ReLu	512 nodes, ReLu	4096 nodes, ReLu	4096 nodes, ReLu
	2	32 nodes, ReLu	32 nodes, ReLu	32 nodes, ReLu	32 nodes, ReLu	4096 nodes, ReLu	4096 nodes, ReLu
Output		1 node	1 node	1 node	1 node	1 node	1 node