Table 2. (a) Alexnet layers architecture and (b) number of selected features after CFS.

(a) Layer type	Size	Kernels size	Number of features
Image input	227 × 227 × 3		150,528
Convolution layer#1	11 × 11 × 3	96	253,440
Activation function
Channel normalization
Pooling
Convolution layer#2	5 × 5 × 48	256	186,624
Activation function
Convolution layer#3	3 × 3 × 256	384	64,896
Activation function
Channel normalization
Pooling
Convolution layer#4	3 × 3 × 192	384	64,896
Activation function
Convolution layer#5	3 × 3 × 192	256	43,264
Activation function
Pooling
Fully connected layer			4,096
Activation function
Dropout
Fully connected layer			4,096
Activation function
Dropout
Fully connected layer			1,000