mnist数据集官网：http://yann.lecun.com/exdb/mnist/
mnist数据集是一个被广泛应用（“嚼烂”）的手写体数字数据集，包含60000个训练样本及10000个测试样本，以字节形式存储。在官网下载到的数据是安装包形式，安装包及其解压后数据形式如下：
我们需要注意的是哪个文件是什么数据集，我将对应关系陈列如下：
t10k-images : 测试图像数据集
t10k-labels ：测试标签数据集
train-images : 训练图像数据集
train-labels : 训练标签数据集
关于标签和图像的对应关系我不在此处表达，因为看到这篇文章的同学们应该都是对数据集有一定了解的同学们。
我在搭建神经网络测试数据的时候，参考了网上很多的代码，也搭建了很多不同的网络，引用mnist数据集的方法也测试了许多次。但也失败了好多，我最终找到了如下方法可以达到预期目标。
首先将mnist数据集转换为CSV格式：
（参考网站：https://blog.csdn.net/Albert201605/article/details/79893585）
我将个人转换代码张贴如下：
def convert(imgf, labelf, outf, n):
f = open(imgf,'rb')
o = open(outf,'w')
l = open(labelf,'rb')
f.read(16)
l.read(8)
images = []
for i in range(n):
image = [ord(l.read(1))]
for j in range(28*28):
image.append(ord(f.read(1)))
images.append(image)
for image in images:
o.write(','.join(str(pix)for pixin image) +'\n')
f.close()
o.close()
l.close()
train_image_path ='E:/College/Graduate_Paper/mnist_test/train-images.idx3-ubyte'
train_label_path ='E:/College/Graduate_Paper/mnist_test/train-labels.idx1-ubyte'
test_image_path ='E:/College/Graduate_Paper/mnist_test/t10k-images.idx3-ubyte'
test_label_path ='E:/College/Graduate_Paper/mnist_test/t10k-labels.idx1-ubyte'
convert( train_image_path , train_label_path ,'E:/College/Graduate_Paper/mnist_test/mnist_train.csv' ,60000 )
convert( test_image_path , test_label_path ,'E:/College/Graduate_Paper/mnist_test/mnist_test.csv' ,10000 )
print('Convert finished!')
转换完成后文件格式如下所示：
在此时，我们依旧无法自然语言方式直接读取测试集内的数据。
其次，将CSV格式的数据集读入神经网络进行训练测试：
（参考网址：https://blog.csdn.net/ebzxw/article/details/81591437）
代码张贴如下：
import numpy
import scipy.special
class neuralNetwork:
def __init__(self, inputnodes, hiddennodes, outputnodes, learningrate):
self.inodes = inputnodes
self.hnodes = hiddennodes
self.onodes = outputnodes
self.lr = learningrate
self.wih = (numpy.random.normal(0.0,pow(self.hnodes, -0.5), (self.hnodes,self.inodes)))#shape (200,784)
self.who = (numpy.random.normal(0.0,pow(self.onodes, -0.5), (self.onodes,self.hnodes)))#shape (10,200)
self.activation_function =lambda x: scipy.special.expit(x)
pass
print('初始化神经网络完成')
def train(self, inputs_list, targets_list):
inputs = numpy.array(inputs_list,ndmin=2).T#shape (784,1)
targets = numpy.array(targets_list,ndmin=2).T#shape (10,1)
hidden_inputs = numpy.dot(self.wih, inputs)#shape (200,1)
hidden_outputs =self.activation_function(hidden_inputs)
final_inputs = numpy.dot(self.who, hidden_outputs)#shape (10,1)
final_outputs =self.activation_function(final_inputs)
output_errors = targets - final_outputs#shape (10,1)
hidden_errors = numpy.dot(self.who.T, output_errors)#shape (200,1)
self.who +=self.lr * numpy.dot((output_errors * final_outputs * (1.0 - final_outputs)),
numpy.transpose(hidden_outputs))
self.wih +=self.lr * numpy.dot((hidden_errors * hidden_outputs * (1.0 - hidden_outputs)),
numpy.transpose(inputs))
pass
print('神经网络训练完成')
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_inputs = numpy.dot(self.who, hidden_outputs)
final_outputs =self.activation_function(final_inputs)
return final_outputs
print('神经网络测试完成')
#设置神经网络初始参数
input_nodes =784 # 28 * 28 = 784
hidden_nodes =200
output_nodes =10
learning_rate =0.1
n = neuralNetwork(input_nodes, hidden_nodes, output_nodes, learning_rate)
print('神经网络参数传入完成')
#训练神经网络
training_data_file =open('E:/College/Graduate_Paper/mnist_test/mnist_train.csv','r')
training_data_list = training_data_file.readlines()
training_data_file.close()
# epochs is the number of times the training data set is used for training
epochs =5
for ein range(epochs):
for recordin training_data_list:
all_values = record.split(',')
inputs = (numpy.asfarray(all_values[1:]) /255.0 *0.99) +0.01
targets = numpy.zeros(output_nodes) +0.01
targets[int(all_values[0])] =0.99
n.train(inputs, targets)
pass
print('%d times train result in the followings:'%e)
test_data_file =open('E:/College/Graduate_Paper/mnist_test/mnist_test.csv','r')
test_data_list = test_data_file.readlines()
test_data_file.close()
scorecard = []
for recordin test_data_list:
all_values = record.split(',')
correct_label =int(all_values[0])
inputs = (numpy.asfarray(all_values[1:]) /255.0 *0.99) +0.01
outputs = n.query(inputs)
label = numpy.argmax(outputs)
if (label == correct_label):
scorecard.append(1)
else:
scorecard.append(0)
pass
scorecard_array = numpy.asarray(scorecard)
print('performance = ', scorecard_array.sum() / scorecard_array.size)
pass
代码运行结果展示如下：
更改参数对神经网络识别正确率影响如下所示：
测试数据仅供参考，转载请注明出处。若有疑问，请私信我（不经常上），看到后会尽快与您讨论。若有侵权，请联系我删除此文。
