手动搭建神经网络，手写数字识别

手动搭建三层神经网络，反向传播，梯度下降

反向传播核心公式推导如下：

标题本次练习数据集采用mnist训练集有6000条数据进行训练，20000条测试机，最终准确率为97%。

链接：https://pan.baidu.com/s/1mg3V4PqKX1EzBpOtzXLD5Q
提取码：6666
–来自百度网盘超级会员V4的分享

#!/usr/bin/python3
# -*- coding: utf-8 -*-
# @Time    : 2023/7/21 14:59
# @Author  : qian
# @FileName: 神经网络.py
# @Software: PyCharm
import numpy.random
import scipy.special


class neuralNetwork():
    def __init__(self,inputnodes,hiddennodes,outputnodes,learningrate):
        self.inodes=inputnodes
        self.hnodes=hiddennodes
        self.onodes=outputnodes

        self.lr=learningrate
        # 链接权重矩阵
        self.wih=numpy.random.rand(self.hnodes,self.inodes)-0.5
        self.who=numpy.random.rand(self.onodes,self.hnodes)-0.5

        # 激活函数
        self.activation_function=lambda x:scipy.special.expit(x)
        pass

    def train(self,inputs_list,targets_list):
        inputs=numpy.array(inputs_list,ndmin=2).T
        targets=numpy.array(targets_list,ndmin=2).T
        hidden_inputs=numpy.dot(self.wih,inputs)
        hidden_outputs=self.activation_function(hidden_inputs)
        #得到预测输出值
        final_input=numpy.dot(self.who,hidden_outputs)
        final_output=self.activation_function(final_input)
        #计算误差
        output_errors=targets-final_output
        hidden_errors=numpy.dot(self.who.T,output_errors)
        self.who+=self.lr*numpy.dot((output_errors*final_output*(1.0-final_output)),numpy.transpose(hidden_outputs))
        self.wih+=self.lr*numpy.dot(hidden_errors*hidden_outputs*(1.0-hidden_outputs),numpy.transpose(inputs))

        pass


    def query(self,inputs_list):
        inputs=numpy.array(inputs_list,ndmin=2).T
        hidden_inputs=numpy.dot(self.wih,inputs)
        hidden_outputs=self.activation_function(hidden_inputs)
        final_intputs=numpy.dot(self.who,hidden_outputs)
        final_outputs=self.activation_function(final_intputs)
        return final_outputs

input_node=784
hidden_nodes=200
output_nodes=10
learning_rate=0.1

n=neuralNetwork(input_node,hidden_nodes,output_nodes,learning_rate)

data_file=open("./mnist_train.csv",'r')
data_list=data_file.readlines()
data_file.close()

test_data_file=open("./mnist_test.csv",'r')
test_data_list=test_data_file.readlines()
test_data_file.close()
for e in range(5):
    for record in data_list:
        all_values = record.split(',')
        inputs = numpy.asfarray(all_values[1:]) / 255.0 * 0.99 + 0.01
        targeys = numpy.zeros(output_nodes) + 0.01
        targeys[int(all_values[0])] = 0.99
        n.train(inputs,targeys)

score=[]
epochs=5

for test_data in test_data_list:
    test_values=test_data.split(',')
    correct_label=int(test_values[0])
    print(correct_label,"正确答案")
    output=n.query((numpy.asfarray(test_values[1:])/255.0*0.99)+0.01)
    label=numpy.argmax(output)
    print(label,"训练结果答案")
    if label==correct_label:
        score.append(1)
    else:
        score.append(0)
print(score)
score_array=numpy.asarray(score)
print("正确率为：",score_array.sum()/score_array.size)
# test_values=test_data_list[0].split(',')
# print(test_values[0])
# print(n.query(numpy.asfarray(test_values[1:])/255.0*0.99)+0.01)