#手写卷积神经网络用于手写数字识别

说明:

想到之前硕士室友秋招找工作面试回来时候曾经和我吐槽，

说面试官让他手写个CNN网络，因为平时直接调用，所以一时间，他也没有写出来。

过了快一年后今天我突然想到了这个问题，所以就写了这个case。

整体是参考知乎大佬链接的。Python 徒手实现 卷积神经网络 CNN - 知乎 (zhihu.com)

撰写部分

卷积层：提取特征

池化层：通过池化减少卷积层的信息冗余

softmax：进行分类

通过交叉熵损失函数计算

进行代码调用。

#Handwritten Convolutional Neural Networks for Handwritten Digit Recognition ##Explanation: I thought that Qiu Zhao, my master's roommate, had roast to me when he came back from his job interview, The interviewer asked him to handwrite a CNN network, but because he usually calls it directly, he didn't write it out for a while. After almost a year, today I suddenly thought of this question, so I wrote this case. The overall reference is based on the links of Zhihu experts. [Python Freehand Implementation of Convolutional Neural Network CNN - Zhihu (zhihu. com)]（ https://zhuanlan.zhihu.com/p/102119808 ） ##Writing section Convolutional layers: extracting features Pooling layer: reducing information redundancy in convolutional layers through pooling Softmax: classify Calculated through cross entropy loss function Make code calls.