TensorFlow基础实验
一、TensorFlow安装
在Windows系统下进行安装，cmd进入命令控制窗，输入命令利用豆瓣网的镜像下载安装TensorFlow包
python -m pip install tensorflow -i https://pypi.douban.com/simple
输入import tensorflow as tf 若不报错，则安装成功。
二、TensorFlow基本操作
3.打开basic-operations.py文件，编写tensorflow基础操作代码。在Python环境，使用import导入TensorFlow模块，别名为tf。
1. import tensorflow as tf
2. import os
3. os.environ["CUDA_VISIBLE_DEVICES"]="0"
4.构造计算图，创建两个常量节点a，b，值分别为2，3，代码如下：
1. a=tf.constant(2)
2. b=tf.constant(3)
5.创建一个Session会话对象，调用run方法，运行计算图。
1. with tf.Session() as sess:
2. print("a:%i" % sess.run(a),"b:%i" % sess.run(b))
3. print("Addition with constants: %i" % sess.run(a+b))
4. print("Multiplication with constant:%i" % sess.run(a*b))
6.代码编写完毕，在basic-operations.py文件内，点击右键=》Run ‘basic-operations’，执行basic-operations.py文件。
7.运行结果为：
8.使用变量Variable构造计算图a，b
1. a=tf.placeholder(tf.int16)
2. b=tf.placeholder(tf.int16)
9.使用tf中的add，multiply函数对a，b进行求和与求积操作。
1. add=tf.add(a,b)
2. mul=tf.multiply(a,b)
10.创建一个Session会话对象，调用run方法，运行计算图。
1. with tf.Session() as sess:
2. print("Addition with variables: %i" % sess.run(add,feed_dict={a:2,b:3}))
3. print("Multiplication with variables: %i" % sess.run(mul,feed_dict={a:2,b:3}))
11.将步骤8，9，10的代码追加到basic-operations.py文件中。运行basic-operations.py文件，运行结果为
12.构造计算图，创建两个矩阵常量节点matrix1，matrix2，值分别为[[3.,3.]]，[[2.],[2.]]，代码如下：
1. matrix1=tf.constant([[3.,3.]])
2. matrix2=tf.constant([[2.],[2.]])
13.构造矩阵乘法运算，
1. product=tf.matmul(matrix1,matrix2)
14.创建一个Session会话对象，调用run方法，运行计算图。
with tf.Session() as sess:
1. result=sess.run(product)
2. print(result)
15.将步骤12，13，14的代码追加basic-operations.py文件中。运行basic_operations.py文件，总的运行结果：
完整代码如下：
import tensorflow as tf import os os.environ["CUDA_VISIBLE_DEVICES"]="0" tf.compat.v1.disable_eager_execution() # a = tf.constant(2) # b = tf.constant(3) # with tf.compat.v1.Session() as sess: # print("a:%i" % sess.run(a), "b:%i" % sess.run(b)) # print("Addition with constants: %i" % sess.run(a + b)) # print("Multiplication with constant:%i" % sess.run(a * b)) # a = tf.compat.v1.placeholder(tf.int16) # b = tf.compat.v1.placeholder(tf.int16) # add = tf.add(a,b) # mul = tf.multiply(a,b) # with tf.compat.v1.Session() as sess: # print("Addition with variables: %i" % sess.run(add,feed_dict={a:2,b:3})) # print("Multiplication with variables: %i" % sess.run(mul, feed_dict={a: 2, b: 3})) matrix1=tf.constant([[3.,3.]]) matrix2=tf.constant([[2.],[2.]]) product=tf.matmul(matrix1,matrix2) with tf.compat.v1.Session() as sess: result=sess.run(product) print(result)
三、TensorFlow线性回归
3.打开linear_regression.py文件，编写tensorflow线性回归代码。导入实验所需要的模块
1. import tensorflow as tf
2. import numpy as np
3. import matplotlib.pyplot as plt
4. import os
5. os.environ["CUDA_VISIBLE_DEVICES"]="0"
4.设置训练参数，learning_rate=0.01，training_epochs=1000，display_step=50。
1. learning_rate=0.01
2. training_epochs=1000
3. display_step=50
5.创建训练数据
1. train_X=np.asarray([3.3,4.4,5.5,6.71,6.93,4.168,9.779,6.182,7.59,2.167,
2. 7.042,10.791,5.313,7.997,5.654,9.27,3.1])
3. train_Y=np.asarray([1.7,2.76,2.09,3.19,1.694,1.573,3.366,2.596,2.53,1.221,
4. 2.827,3.465,1.65,2.904,2.42,2.94,1.3])
5. n_samples=train_X.shape[0]
6.构造计算图，使用变量Variable构造变量X，Y，代码如下：
1. X=tf.placeholder("float")
2. Y=tf.placeholder("float")
7.设置模型的初始权重
1. W=tf.Variable(np.random.randn(),name="weight")
2. b=tf.Variable(np.random.randn(),name='bias')
8.构造线性回归模型
1. pred=tf.add(tf.multiply(X,W),b)
9.求损失函数，即均方差
1. cost=tf.reduce_sum(tf.pow(pred-Y,2))/(2*n_samples)
10.使用梯度下降法求最小值，即最优解
1. optimizer=tf.train.GradientDescentOptimizer(learning_rate).minimize(cost)
11.初始化全部变量
1. init =tf.global_variables_initializer()
12.使用tf.Session()创建Session会话对象，会话封装了Tensorflow运行时的状态和控制。
1. with tf.Session() as sess:
2. sess.run(init)
13.调用会话对象sess的run方法，运行计算图，即开始训练模型。
1. #Fit all training data
2. for epoch in range(training_epochs):
3. for (x,y) in zip(train_X,train_Y):
4. sess.run(optimizer,feed_dict={X:x,Y:y})
5.
6. #Display logs per epoch step
7. if (epoch+1) % display_step==0:
8. c=sess.run(cost,feed_dict={X:train_X,Y:train_Y})
9. print("Epoch:" ,'%04d' %(epoch+1),"cost=","{:.9f}".format(c),"W=",sess.run(W),"b=",sess.run(b))
14.打印训练模型的代价函数。
1. training_cost=sess.run(cost,feed_dict={X:train_X,Y:train_Y})
2. print("Train cost=",training_cost,"W=",sess.run(W),"b=",sess.run(b))
15.可视化，展现线性模型的最终结果。
1. plt.plot(train_X,train_Y,'ro',label='Original data')
2. plt.plot(train_X,sess.run(W)*train_X+sess.run(b),label="Fitting line")
3. plt.legend()
4. plt.show()
16.完整代码如下：
1. import tensorflow as tf
2. import numpy as np
3. import matplotlib.pyplot as plt
4. import os
5. os.environ["CUDA_VISIBLE_DEVICES"]="0"
6. #Parameters
7. learning_rate=0.01
8. training_epochs=1000
9. display_step=50
10. #training Data
11. train_X=np.asarray([3.3,4.4,5.5,6.71,6.93,4.168,9.779,6.182,7.59,2.167,
12. 7.042,10.791,5.313,7.997,5.654,9.27,3.1])
13. train_Y=np.asarray([1.7,2.76,2.09,3.19,1.694,1.573,3.366,2.596,2.53,1.221,
14. 2.827,3.465,1.65,2.904,2.42,2.94,1.3])
15. n_samples=train_X.shape[0]
16. #tf Graph Input
17. X=tf.placeholder("float")
18. Y=tf.placeholder("float")
19. #Set model weights
20. W=tf.Variable(np.random.randn(),name="weight")
21. b=tf.Variable(np.random.randn(),name='bias')
22. #Construct a linear model
23. pred=tf.add(tf.multiply(X,W),b)
24. #Mean squared error
25. cost=tf.reduce_sum(tf.pow(pred-Y,2))/(2*n_samples)
26. # Gradient descent
27. optimizer=tf.train.GradientDescentOptimizer(learning_rate).minimize(cost)
28. #Initialize the variables
29. init =tf.global_variables_initializer()
30. #Start training
31. with tf.Session() as sess:
32. sess.run(init)
33. #Fit all training data
34. for epoch in range(training_epochs):
35. for (x,y) in zip(train_X,train_Y):
36. sess.run(optimizer,feed_dict={X:x,Y:y})
37. #Display logs per epoch step
38. if (epoch+1) % display_step==0:
39. c=sess.run(cost,feed_dict={X:train_X,Y:train_Y})
40. print("Epoch:" ,'%04d' %(epoch+1),"cost=","{:.9f}".format(c),"W=",sess.run(W),"b=",sess.run(b))
41. print("Optimization Finished!")
42. training_cost=sess.run(cost,feed_dict={X:train_X,Y:train_Y})
43. print("Train cost=",training_cost,"W=",sess.run(W),"b=",sess.run(b))
44. #Graphic display
45. plt.plot(train_X,train_Y,'ro',label='Original data')
46. plt.plot(train_X,sess.run(W)*train_X+sess.run(b),label="Fitting line")
47. plt.legend()
48. plt.show()
17.代码编写完毕，在linear_regression.py文件内，点击右键=》Run ‘linear_regression’，执行linear_regression.py文件。
18.运行结果为：
完整代码如下：
import tensorflow as tf import numpy as np import matplotlib.pyplot as plt import os os.environ["CUDA_VISIBLE_DEVICES"]="0" tf.compat.v1.disable_eager_execution() learning_rate=0.01 training_epochs=1000 display_step=50 train_X=np.asarray([3.3,4.4,5.5,6.71,6.93,4.168,9.779,6.182,7.59,2.167,7.042,10.791,5.313,7.997,5.654,9.27,3.1]) train_Y=np.asarray([1.7,2.76,2.09,3.19,1.694,1.573,3.366,2.596,2.53,1.221,2.827,3.465,1.65,2.904,2.42,2.94,1.3]) n_samples=train_X.shape[0] X=tf.compat.v1.placeholder("float") Y=tf.compat.v1.placeholder("float") W=tf.Variable(np.random.randn(),name="weight") b=tf.Variable(np.random.randn(),name="bias") pred=tf.add(tf.multiply(X,W),b) cost=tf.reduce_sum(tf.pow(pred-Y,2))/(2*n_samples) optimizer=tf.compat.v1.train.GradientDescentOptimizer(learning_rate).minimize(cost) init=tf.compat.v1.global_variables_initializer() with tf.compat.v1.Session() as sess: sess.run(init) #Fit all training data for epoch in range(training_epochs): for (x,y) in zip(train_X,train_Y): sess.run(optimizer,feed_dict={X:x,Y:y}) #Display logs per epoch step if (epoch+1) % display_step==0: c=sess.run(cost,feed_dict={X:train_X,Y:train_Y}) print("Epoch:" ,'%04d' %(epoch+1),"cost=","{:.9f}".format(c),"W=",sess.run(W),"b=",sess.run(b)) print("Optimization Finished!") training_cost=sess.run(cost,feed_dict={X:train_X,Y:train_Y}) print("Train cost=",training_cost,"W=",sess.run(W),"b=",sess.run(b)) #Graphic display plt.plot(train_X, train_Y, 'ro', label='Original data') plt.plot(train_X, sess.run(W) * train_X + sess.run(b), label="Fitting line") plt.legend() plt.show()
