黄伟Tensorflow笔记一

Tensorflow笔记一

一. 目的：

生成干扰数据 计算 Y = 2 * X   的结果

二.代码

# 完整代码
# 计算 Y = 2 * X    的值
import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt

def moving_average(a,w=10):
    if(len(a)<w):
        return a[:]
    return [val if idx<w else sum(a[(idx-w):idx])/w for idx,val in enumerate(a)]



# 生成100 个 -1 到 1 的数据
trainX = np.linspace(-1,1,100)

# 生成100个干扰数据
trainY = 2 * trainX + np.random.randn(*trainX.shape) * 0.3

# 绘图
plt.plot(trainX,trainY,'ro',label = 'I Love China')
plt.legend()
plt.show()

# 占位
X = tf.placeholder("float")
Y = tf.placeholder("float")

# 模型参数
W = tf.Variable(tf.random_normal([1]), name = 'bias')
B = tf.Variable(tf.zeros([1]),name="bias")


# 前向结构   前向传播
Z = tf.multiply(X,W) + B


#  反向优化
cost = tf.reduce_mean(tf.square(Y - Z))
learning_rate = 0.01
optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(cost)


# 迭代训练模型
init = tf.global_variables_initializer()

# 定义参数
training_epochs = 20
display_step = 2

#  启动session
with tf.Session() as sess:
    sess.run(init)
    plotdata = {"batchsize":[],"loss":[]}
    
    # 向模型输入数据
    for epoch in range(training_epochs):
        for(x,y) in zip(trainX,trainY):
           sess.run(optimizer,feed_dict={X:x,Y:y})
        
        # 显示训练中的详细信息
        if epoch % display_step == 0:
            loss = sess.run(cost,feed_dict={X:trainX,Y: trainY})
            print("Epoch:",epoch + 1,"cost = ",loss,"W=",sess.run(W),"b=",sess.run(B))
            if not (loss == "NA"):
                plotdata["batchsize"].append(epoch)
                plotdata["loss"].append(loss)
                
                
                
                
    print("Finished")
    print("cost=",sess.run(cost,feed_dict={X:trainX,Y: trainY}),"W=",sess.run(W),"b=",sess.run(B))
    
    # 图像显示  
    plt.plot(trainX,trainY,'ro',label = 'sad data')
    plt.plot(trainX,sess.run(W)*trainX + sess.run(B),label='Fittedline')
    plt.legend()
    plt.show()
    
    
    plotdata["avgloss"] = moving_average(plotdata["loss"])
    plt.figure(1)
    plt.subplot(211)
    plt.plot(plotdata["batchsize"],plotdata["avgloss"],'b--')
    plt.xlabel('minibatch number')
    plt.ylabel('Loss')
    plt.title('Minibatch run vs. Training loss')
    plt.show()
    
    print("x=0.2,z=",sess.run(Z,feed_dict={X:0.2}))

三、运行结果

Epoch: 1 cost = 0.19036867 W= [1.5034635] b= [0.15345636]
Epoch: 3 cost = 0.09338717 W= [1.8725581] b= [0.05529383]
Epoch: 5 cost = 0.08592511 W= [1.9727246] b= [0.01758706]
Epoch: 7 cost = 0.085566826 W= [1.9987036] b= [0.00763]
Epoch: 9 cost = 0.08558113 W= [2.0054226] b= [0.00505186]
Epoch: 11 cost = 0.08559199 W= [2.00716] b= [0.00438513]
Epoch: 13 cost = 0.08559528 W= [2.00761] b= [0.00421249]
Epoch: 15 cost = 0.085596174 W= [2.0077255] b= [0.00416822]
Epoch: 17 cost = 0.08559638 W= [2.0077548] b= [0.00415687]
Epoch: 19 cost = 0.08559645 W= [2.0077631] b= [0.00415374]
Finished
cost= 0.08559646 W= [2.0077643] b= [0.00415323]

x=0.2,z= [0.4057061]

x=0.2,z= [0.4057061]

四、小结

（1）linspace

# 作用：生成100 个 -1 到 1的数 
# start  参数1： 起始位置
# stop   参数2： 结束位置
# num    参数3： 生成个数，默认50
# endpoint  参数4 ：是否包含 stop 数值，默认为True，包含stop值；若为False，则不包含stop值
# retstep   参数5:  默认为False，返回等差数列数组[-1.  -0.5  0.   0.5  1. ]，
#                  若为True，则返回结果(array([`samples`, `step`]))
# return 返回： num 个 start 到 stop  之间数组成的数组或列表

trainX = np.linspace(-1,1,100)