Symbol	Description
q	Dropout rate
lr	Batch size
m	Learning rate
ke	Convolution kernel size (kernel height = kernel width)
K	Number of sample categories
X	Sample dataset
${\mathrm{z}}^{(l)}$	Vector input to the l layer
$y^{(l)}$	Output vector from the l layer
$f(z)$	Activation function
$x_i^{(l)}$	Input of the l layer
$w_i^{(l)}$	Weight of the l layer of $x_i^{(l)}$
$b_i^{(l)}$	Bias of the l layer of $x_i^{(l)}$
${\overrightarrow{x}}_{i,k}^{(l)}$	Input of the sample of the K category in the l layer
$w_{k,i}^{(l)}$	Weight of the l layer of class k
$b_{k,i}^{(l)}$	Bias of the l layer of class k
$I$	Input image size
$S$	Stride size
$P$	Padding size
$O$	Output image size
$W_c$	Number of weights the convolutional layer
$B_c$	Number of biases of the convolutional layer
$P_c$	Number of all parameters of the convolutional layer
$N$	Set number of cores
$CH$	Number of input image channels
$W_{cf}$	Number of biases of the fully connected layer connected to the convolutional layer
$B_{cf}$	Number of weights of the fully connected layer connected to the convolutional layer
$o^{\prime }$	Output image size of the previous convolution layer
$P_{cf}$	Total number of parameters in the current fully connected layer
F	Number of neurons in the current fully connected layer
F −1	Number of neurons in the previous fully connected layer