从《python神经网络编程》一书中提取的识别手写字体的神经网络代码
训练集:http://www.pjreddie.com/media/files/mnist_train.csv
测试集:http://www.pjreddie.com/media/files/mnist_test.csv
import numpy
# scipy.special for the sigmoid function expit()
import scipy.special
import matplotlib.pyplot
from matplotlib import pyplot as plt
%matplotlib inline
class neuralNetwork :
def __init__(self, inputnodes, hiddennodes, outputnodes,learningrate) :
# 定义输入层。隐藏层和输出层的节点数
self.inodes = inputnodes
self.hnodes = hiddennodes
self.onodes = outputnodes
# 定义,系数矩阵,服从高斯分布的概率密度函数(正态分布),numpy.random.normal(loc=0.0, scale=1.0, size=None)
# loc:float概率分布的均值,对应着整个分布的中心center
# scale:float概率分布的标准差,对应于分布的宽度,scale越大越矮胖,scale越小,越瘦高
# size:int or tuple of ints输出的shape,默认为None,只输出一个值我们更经常会用到
# np.random.randn(size)所谓标准正太分布(μ=0, σ=1),对应于np.random.normal(loc=0, scale=1, size)
# (self.onodes, self.hnodes)是得到的数组形状,pow(self.onodes, -0.5)对self.onodes进行-0.2开方
self.wih = numpy.random.normal(0.0, pow(self.hnodes, -0.5),(self.hnodes, self.inodes))
self.who = numpy.random.normal(0.0, pow(self.onodes, -0.5),(self.onodes, self.hnodes))
# self.lr是学习率
self.lr = learningrate
# 激活函数sigmoid
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)
final_inputs = numpy.dot(self.who, hidden_outputs)
final_outputs = self.activation_function(final_inputs)
# 计算损失矩阵,目标值-实际输出值
output_errors = targets - final_outputs
# 隐藏层的损失矩阵
hidden_errors = numpy.dot(self.who.T, output_errors)
#更新权重矩阵
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))
# 返回输出结果
def query(self, inputs_list) :
inputs = numpy.array(inputs_list, ndmin=2).T
# 权重矩阵(系数矩阵)和输入进行点乘,得到隐藏层节点的数据矩阵
hidden_inputs = numpy.dot(self.wih, inputs)
# 将得到的矩阵作用于激活函数,得到隐藏层节点的输出矩阵
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
pass
input_nodes = 784
hidden_nodes = 100
output_nodes = 10
learning_rate = 0.2
n = neuralNetwork(input_nodes,hidden_nodes,output_nodes,learning_rate)
training_data_file= open("F:/matlabCode/hands datasets/mnist_train.csv", 'r')
training_data_list= training_data_file.readlines()
training_data_file.close()
for record in 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
test_data_file=open("F:/matlabCode/hands datasets/mnist_test.csv", 'r')
test_data_list=test_data_file.readlines()
test_data_file.close()
all_values=test_data_list[0].split(',')
scorecard = []
for record in test_data_list:
# record是字符串,split将字符串进行切割并以list存储
all_values = record.split(',')
# 将列表第一个元素得到,即标签值,并将其强制转换为int型
correct_label = int(all_values[0])
# print(correct_label, "correct label")
# 输入值是0-255,将其范围进行转换,使其范围在0.01-0.99之间
inputs = (numpy.asfarray(all_values[1:]) / 255.0 * 0.99) + 0.01
outputs = n.query(inputs)
label = numpy.argmax(outputs)
# print(label, "network's answer")
if (label == correct_label):
scorecard.append(1)
else:
scorecard.append(0)
pass
pass
scorecard_array = numpy.asarray(scorecard)
print(scorecard_array.sum())
# 输出成功率
print ("performance = ", scorecard_array.sum() /scorecard_array.size)
这段代码输出结果为成功率%
9448
performance = 0.9448
import numpy
import matplotlib.pyplot
%matplotlib inline
# all_values= training_data_list[0].split(',')
image_array= numpy.asfarray(all_values[1:]).reshape((28,28))
# 将值转换在0到1之间
# scaled_input=(numpy.asfarray(all_values[1:])/255.0*0.99+0.01)
matplotlib.pyplot.imshow( image_array, cmap='Greys',interpolation=None)
# input_nodes=3
# hidden_nodes=3
# out_puts=3
# learning_rate=0.3
# n=neuralNetwork(input_nodes,hidden_nodes,out_puts,learning_rate)
# n.query([1.0,0.5,-1.5])
输出为具体的数字