本文共 3355 字，大约阅读时间需要 11 分钟。

1. 纯python写的最简单的三层神经网络，只有输入层、一个隐藏层、输出层，使MNIST100条数据进行训练，10条数据进行测试。
2. 测试环境Win10 Python3.7.5 pyCharm
3. 100条训练数据：
4. 10条测试数据：
5. 神经网络可视化学习工具TensorFlow Playground：

import numpyimport scipy.specialimport matplotlib.pyplotimport pylab#神经网络类class neuralNetwork:    def __init__(self, inputnodes, hiddennodes, outputnodes,learningrate):        # inputnodes输入节点数目，即图像的宽*高=28*28=784        self.inodes = inputnodes        self.hnodes= hiddennodes        self.onodes = outputnodes        #初始化权重服从正态分布        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 = learningrate;        #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        #正向推算        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)        #误差反向推算        output_errors = targets - final_outputs        self.who += self.lr * numpy.dot((output_errors * final_outputs * (1.0 - final_outputs)), numpy.transpose(hidden_outputs))        hidden_errors = numpy.dot(self.who.T, output_errors)        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_inputs = numpy.dot(self.who, hidden_outputs)        final_outputs = self.activation_function(final_inputs)        return final_outputs#训练神经网络input_nodes = 784hidden_nodes = 100output_nodes = 10learning_rate = 0.05n = neuralNetwork(input_nodes, hidden_nodes, output_nodes, learning_rate)training_data_file = open("mnist_train_100.csv", 'r')training_data_list = training_data_file.readlines()training_data_file.close()epochs = 20for e in range(epochs):    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    pass#测试神经网络test_data_file = open("mnist_test_10.csv", 'r')test_data_list = test_data_file.readlines()test_data_file.close()scorecard = []for record in test_data_list:    record_value = record.split(',')    image_array = numpy.asfarray(record_value[1:]).reshape(28, 28)    matplotlib.pyplot.imshow(image_array, cmap="Greys", interpolation='None')    pylab.show()    inputs = (numpy.asfarray(record_value[1:]) / 255.0 * 0.99) + 0.01    correct_label = int(record_value[0])    outputs = n.query(inputs)    label = numpy.argmax(outputs)    if(label == correct_label):        print("right:", label)        scorecard.append(1)    else:        print("wrong:", label,"v", correct_label)        scorecard.append(0)        pass    passscorecard_array = numpy.asarray(scorecard)print("performance= ", scorecard_array.sum() / scorecard_array.size)

转载地址：https://blog.csdn.net/zouxin_88/article/details/103210254 如侵犯您的版权，请留言回复原文章的地址，我们会给您删除此文章，给您带来不便请您谅解！