1、什么是tensorflow
tensorflow是一个开源软件库，使用data flow graphs进行数值计算，最初由Google大脑团队开发，用于机器学习和深度卷积网络的研究，同样适用于其他广泛的领域。
2、访问tensorflow官网：在Windows的host文件（位置在C:\Windows\System32\drivers\etc\hosts）末尾添加如下内容：
#TensorFlow start
64.233.188.121 www.tensorflow.org
#TensorFlow end
打开cmd，输入如下命令：ipconfig /flushdns 接着访问www.tensorflow.org，就能打开主页了。
3、例子一
import tensorflow as tf #载入tensorflow库
m1 = tf.constant([[1, 2]]) #op(节点)
m2 = tf.constant([[3],[4]]) #op(节点)
m1xm2 = tf.matmul(m1, m2) #op(节点)
# method1 use session
sess = tf.Session()
result = sess.run(m1xm2)
print(result)
sess.close()
# method2 use session, 在被称之为 会话 (Session) 的上下文 (context) 中执行图，此处默认图
with tf.Session() as sess:
result = sess.run(m1xm2)
print(result)
4、例子二
import tensorflow as tf #载入tensorflow库
x1 = tf.placeholder(dtype = tf.float32) #占位符，运行的时候通过feed_dict参数传递
y1 = tf.placeholder(dtype=tf.float32) #占位符，运行的时候通过feed_dict参数传递
z1 = x1 + y1
x2 = tf.placeholder(dtype = tf.float32, shape = [2,3])
y2 = tf.placeholder(dtype = tf.float32, shape = [3,2])
z2 = tf.matmul(x2, y2)
with tf.Session() as sess:
z1_out = sess.run(z1, feed_dict = {x1:2, y1:4}) # 通过feed_dict参数传递，x1=2, y1=4
print(z1_out)
z2_out= sess.run(z2, feed_dict = {x2 : [[1,2,3], [4,5,6]], y2 : [[1,2],[3,4],[5,6]]})
print (z2_out)
5、例子三
import tensorflow as tf #载入tensorflow库
v1 = tf.Variable(1)
addTest = tf.add(v1,2)
update_addTest = tf.assign(v1, addTest) #将addTestf赋值给v1
with tf.Session() as sess:
#变量必须要初始化
sess.run(tf.global_variables_initializer())
for i in range(5):
sess.run(update_addTest)
print(sess.run(v1))
6、例子四
import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
# fake data
x = np.linspace(-10, 10, 100) # [-10,10],共100个点
# following are popular activation functions
y_relu = tf.nn.relu(x) #relu
y_sigmoid = tf.nn.sigmoid(x) #sigmoid
y_tanh = tf.nn.tanh(x) #tanh
y_softplus = tf.nn.softplus(x) #softplus
# y_softmax = tf.nn.softmax(x) softmax is a special kind of activation function, it is about probability
with tf.Session() as sess:
y_relu, y_sigmoid, y_tanh, y_softplus = sess.run([y_relu, y_sigmoid, y_tanh, y_softplus])
# plt to visualize these activation function
plt.figure(1, figsize=(8, 6)) #figure1，宽度8，高度6
plt.subplot(221)
plt.plot(x, y_relu, c='red', label='relu')#红色线，左上角relu
plt.ylim((-2, 5))
plt.legend(loc='best')
plt.subplot(222)
plt.plot(x, y_sigmoid, c='red', label='sigmoid')
plt.ylim((-0.5, 1))
plt.legend(loc='best')
plt.subplot(223)
plt.plot(x, y_tanh, c='red', label='tanh')
plt.ylim((-1.2, 1.2))
plt.legend(loc='best')
plt.subplot(224)
plt.plot(x, y_softplus, c='red', label='softplus')
plt.ylim((-0.2, 6))
plt.legend(loc='best')
plt.show()
7、例子五
import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
#数据
x = np.linspace(-1,1,100)[:,np.newaxis] #列向量，100行1列
noise = np.random.normal(0,0.1,size = x.shape) #高斯噪声，均值0，方差0.1
y = np.power(x,2) + noise #x的平方+noise
plt.scatter(x,y) #散点图
plt.show()
tf_x = tf.placeholder(tf.float32, x.shape)
tf_y = tf.placeholder(tf.float32, y.shape)
hid = tf.layers.dense(tf_x, 10, tf.nn.relu)
output = tf.layers.dense(hid, 1)
loss = tf.losses.mean_squared_error(tf_y, output)
optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.1)
train_op = optimizer.minimize(loss)
sess = tf.Session()
sess.run(tf.global_variables_initializer())
plt.ion()
for step in range(500):
_,l,predict = sess.run([train_op,loss,output], {tf_x:x, tf_y:y})
if step % 10 == 0:
plt.cla()
plt.scatter(x,y)
plt.plot(x,predict,'r-',lw=5)
plt.text(0.5, 0, 'loss=%0.4f'% l, fontdict={'size':20, 'color':'red'})
plt.pause(0.1)
plt.ioff()
plt.show()
sess.close()
print('end...')
8、例子六