由于特征向量的归一化操作不涉及参数学习，所以本章节的伪代码部分不特别写

Since the normalization operation of the feature vector does not involve parameter learning, the pseudo-code part of this chapter is not specifically written, and the normalization operation of the feature vector is generally incorporated into the convolutional network module. The neural network training process of the mixed loss function of Softmax loss and centrality loss is shown in Pseudocode Algorithm 1.

Since the normalization operation of the feature vector does not involve parameter learning, the pseudo-code part of this chapter is not specifically written, and the normalization operation of the feature vector is generally included in the reel network module. The neural network training process for softmax loss and central loss hybrid loss functions is shown in pseudo-code algorithm 1.

Since the normalization operation of eigenvector does not involve parameter learning, the pseudo code part of this chapter is not specially written, and the normalization operation of eigenvector is generally incorporated into convolution network module. The neural network training process of softmax loss and centrality loss mixed loss function is shown in pseudo code algorithm 1.<br>