diff --git "a/31-\346\264\252\344\271\205\350\277\234 \345\214\227\344\272\254/\347\254\254\344\271\235\345\221\250/Keras2.py" "b/31-\346\264\252\344\271\205\350\277\234 \345\214\227\344\272\254/\347\254\254\344\271\235\345\221\250/Keras2.py"
new file mode 100644
index 00000000..4a186393
--- /dev/null
+++ "b/31-\346\264\252\344\271\205\350\277\234 \345\214\227\344\272\254/\347\254\254\344\271\235\345\221\250/Keras2.py"	
@@ -0,0 +1,53 @@
+# import tensorflow as tf
+from tensorflow.python.keras.datasets import mnist
+import matplotlib.pyplot as plt
+# import cv2
+from tensorflow.python.keras import models
+from tensorflow.python.keras import layers
+from tensorflow.python.keras.utils import to_categorical
+
+
+# 1. 导入数据并查看
+
+(trainImages,trainLables),(testImages,testLables)=mnist.load_data()
+# print('trainImages.shape:',trainImages.shape)
+# print('trainLables:',trainLables)
+# print('testImages.shape:',testImages.shape)
+# print('testLables:',testLables)
+#查看测试集第一张图片
+# test0=testImages[0]
+# plt.imshow(test0,cmap=plt.cm.binary)
+# plt.show()
+#cv2.imshow('test0',test0)
+#cv显示较小
+
+
+
+
+#2. 使用keras构建神经网络
+
+net=models.Sequential()
+net.add(layers.Dense(units=512,activation='relu',input_shape=(28*28,)))
+net.add(layers.Dense(units=10,activation='softmax'))
+net.compile(optimizer='rmsprop',loss='categorical_crossentropy',metrics=['accuracy'])
+
+#数据重塑与归一化
+trainImages=trainImages.reshape((60000,28*28))
+trainImages=trainImages.astype('float32')/255
+testImages=testImages.reshape((10000,28*28))
+testImages=testImages.astype('float32')/255
+#标签独热化
+testLablesOriginal=testLables
+trainLables=to_categorical(trainLables)
+testLables=to_categorical(testLables)
+
+#3. 训练与测试
+
+net.fit(trainImages,trainLables,epochs=7,batch_size=128)
+
+testLoss,testAccuracy=net.evaluate(testImages,testLables,verbose=1)
+print('\nTest Loss:',testLoss,'Test Accuracy:',testAccuracy)
+
+result=net.predict(testImages)
+
+
diff --git "a/31-\346\264\252\344\271\205\350\277\234 \345\214\227\344\272\254/\347\254\254\344\271\235\345\221\250/NeuralNetWork2.py" "b/31-\346\264\252\344\271\205\350\277\234 \345\214\227\344\272\254/\347\254\254\344\271\235\345\221\250/NeuralNetWork2.py"
new file mode 100644
index 00000000..df10229a
--- /dev/null
+++ "b/31-\346\264\252\344\271\205\350\277\234 \345\214\227\344\272\254/\347\254\254\344\271\235\345\221\250/NeuralNetWork2.py"	
@@ -0,0 +1,88 @@
+import numpy as np
+import scipy.special
+
+
+class NeuralNetWork:
+    def __init__(self, inputNodes, hiddenNodes, outputNodes, learningRate):
+        self.iNodes = inputNodes
+        self.hNodes = hiddenNodes
+        self.oNodes = outputNodes
+        self.lr = learningRate
+        # 以上为基本参数初始化
+        self.wih = (np.random.normal(loc=0.0, scale=pow(self.hNodes, -0.5), size=(self.hNodes, self.iNodes)))
+        self.who = (np.random.normal(loc=0.0, scale=pow(self.oNodes, -0.5), size=(self.oNodes, self.hNodes)))
+        # 以上为权重矩阵初始化，wih为输入-隐藏层权重矩阵，who为隐藏-输出权重矩阵
+        # loc为均值，scale为标准差
+        self.activationFunction = lambda x: scipy.special.expit(x)
+        # 以上为激活函数sigmoid
+
+    def train(self, inputList, targetList):
+        inputs = np.array(inputList, ndmin=2).T
+        targets = np.array(targetList, ndmin=2).T
+        hiddenInputs = np.dot(self.wih, inputs)
+        # 以上为隐藏层的输入
+        hiddenOutputs = self.activationFunction(hiddenInputs)
+        #  以上为隐藏层经过激活后的输出
+        finalInputs = np.dot(self.who, hiddenOutputs)
+        # 以上为输出层的输入
+        finalOutputs = self.activationFunction(finalInputs)
+        # 以上为最终输出
+        outputErrors = targets - finalOutputs
+        hiddenErrors = np.dot(self.who.T, outputErrors * finalOutputs * (1 - finalOutputs))
+        self.who += self.lr * np.dot((outputErrors * finalOutputs * (1 - finalOutputs)), np.transpose(hiddenOutputs))
+        self.wih += self.lr * np.dot((hiddenErrors*hiddenOutputs*(1-hiddenOutputs)),np.transpose(inputs))
+
+
+    def query(self,inputs):
+        hiddenInputs=np.dot(self.wih,inputs)
+        hiddenOutputs = self.activationFunction(hiddenInputs)
+        finalInputs = np.dot(self.who, hiddenOutputs)
+        finalOutputs = self.activationFunction(finalInputs)
+        print(finalOutputs)
+        return finalOutputs
+
+
+inputNodes=784
+hiddenNodes=200
+outputNodes=10
+learningRate=0.1
+n=NeuralNetWork(inputNodes,hiddenNodes,outputNodes,learningRate)
+
+#以下读取训练数据
+trainDataFile=open("dataset/mnist_train.csv",'r')
+trainDataList=trainDataFile.readlines()
+trainDataFile.close()
+
+#加入epochs，即循环次数
+
+epochs=5
+for e in range(epochs):
+    for record in trainDataList:
+        allValues=record.split(',')
+        inputs=(np.asfarray(allValues[1:]))/255.0*0.99+0.01
+        #此为图片与数值的对应关系
+        targets=np.zeros(outputNodes)+0.01
+        targets[int(allValues[0])]=0.99
+        n.train(inputs,targets)
+
+testDataFile=open("dataset/mnist_test.csv")
+testDataList=testDataFile.readlines()
+testDataFile.close()
+
+scores=[]
+for record in testDataList:
+    allValuesTest=record.split(',')
+    correctNumber=int(allValuesTest[0])
+    print("该图片对应数字为：",correctNumber)
+    inputs=(np.asfarray(allValuesTest[1:]))/255.0*0.99+0.01
+    outputs=n.query(inputs)
+    label=np.argmax(outputs)
+    print("网络认为的图片为：",label)
+    if label==correctNumber:
+        scores.append(1)
+    else:
+        scores.append(0)
+print(scores)
+
+scoresArray=np.asarray(scores)
+print('performance=',scoresArray.sum()/scoresArray.size)
diff --git "a/31-\346\264\252\344\271\205\350\277\234 \345\214\227\344\272\254/\347\254\254\344\271\235\345\221\250/tf2.py" "b/31-\346\264\252\344\271\205\350\277\234 \345\214\227\344\272\254/\347\254\254\344\271\235\345\221\250/tf2.py"
new file mode 100644
index 00000000..9f20ed68
--- /dev/null
+++ "b/31-\346\264\252\344\271\205\350\277\234 \345\214\227\344\272\254/\347\254\254\344\271\235\345\221\250/tf2.py"	
@@ -0,0 +1,50 @@
+import tensorflow as tf
+import numpy as np
+import matplotlib.pyplot as plt
+
+#使用np生成200随机点
+xData=np.linspace(start=-0.5,stop=0.5,num=200)[:,np.newaxis]
+#横向变纵向
+print(xData.shape)
+noise=np.random.normal(loc=0,scale=0.02,size=xData.shape)
+yData=np.square(xData)+noise
+#建立y=x^2函数并加上噪音
+
+#定义两个placeholder存放输入数据
+#shape：数据形状。默认是None，就是一维值，
+# 也可以是多维,比如[2,3], [None, 3]表示列是3，行不定）
+x=tf.placeholder(dtype=tf.float32,shape=[None,1])
+y=tf.placeholder(dtype=tf.float32,shape=[None,1])
+
+#定义中间层
+#权重使用正态初始化，形状为一行10列
+weightsL1=tf.Variable(tf.random_normal([1, 20]))
+print(weightsL1)
+biasesL1=tf.Variable(tf.zeros([1,20]))
+outputL1= tf.matmul(x, weightsL1) + biasesL1
+print(outputL1)
+L1=tf.nn.tanh(outputL1)
+
+#定义输出层
+weightsL2=tf.Variable(tf.random_normal([20,1]))
+biasesL2=tf.Variable(tf.zeros([1,1]))
+outputL2=tf.matmul(L1,weightsL2)+biasesL2
+prediction=tf.nn.tanh(outputL2)
+
+#定义损失函数为均方差,
+loss=tf.reduce_mean(tf.square(y-prediction))
+#使用学习率为0.1的反向传播算法最小化loss
+trainStep=tf.train.GradientDescentOptimizer(learning_rate=0.1).minimize(loss)
+
+with tf.Session() as sess:
+    sess.run(tf.global_variables_initializer())
+    #全局变量初始化
+    for i in range(1000):
+        sess.run(trainStep,feed_dict={x:xData,y:yData})
+
+    predictValues=sess.run(prediction,feed_dict={x:xData})
+
+    plt.figure()
+    plt.scatter(xData,yData)
+    plt.plot(xData,predictValues,'r-',lw=4)
+    plt.show()
\ No newline at end of file