使用tensorflow训练一个最简单的神经网路

一个最简单的tensorflow+neural network demo

重点是初步理解tensorflow的思想：
最重要的是建立一个网络模型，训练这个事情，可以直接用tf来帮你

import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt

#添加层的函数，层有输入参数，输出参数，偏置，还有激活函数
def add_layer(inputs,in_size,out_size,activation_function=None):
	#随机初始化参数和偏置
    Weights = tf.Variable(tf.random_normal([in_size,out_size]))
    biases = tf.Variable(tf.zeros([1,out_size])+0.1)
    #计算（W*a+b）
    Wx_plus_b = tf.matmul(inputs,Weights)+biases
    #如果有激励函数就A（W*a+b）
    if activation_function is None:
        outputs = Wx_plus_b
    else:
        outputs = activation_function(Wx_plus_b)
    return outputs
#生成训练数据
x_data = np.linspace(-1,1,300)[:,np.newaxis]
noise = np.random.normal(0,0.05,x_data.shape)
y_data = np.square(x_data)-0.5 + noise

#定义输入输出
xs=tf.placeholder(tf.float32,[None,1])
ys=tf.placeholder(tf.float32,[None,1])

#生成l1层，1输入，19输出，relu激活
l1 = add_layer(xs,1,10,activation_function=tf.nn.relu)
#生成prediction层，10输入，1输出，输出直接进行预测
prediction = add_layer(l1,10,1,activation_function=None)

#定义loss函数，均方根误差
loss = tf.reduce_mean(tf.reduce_sum(tf.square(ys-prediction),
                                    reduction_indices=[1]))
#定义训练方式：梯度下降法，误差最小化
train_step = tf.train.GradientDescentOptimizer(0.1).minimize(loss)
#初始化变量
init = tf.initialize_all_variables()
#建立会话
sess = tf.Session()
#初始化
sess.run(init)

#绘图模块
fig = plt.figure()
ax = fig.add_subplot(1,1,1)
ax.scatter(x_data,y_data)
plt.ion()

plt.show()
#进行一千次训练
for i in range(1000):
	#凡是使用到了placeholder的地方都需要feed_dict={xx:xxx,yy:yyy}
    sess.run(train_step,feed_dict={xs:x_data,ys:y_data})
    if i % 50 == 0:
        #print(sess.run(loss,feed_dict={xs:x_data,ys:y_data}))
        #绘图
        try:
            ax.lines.remove(lines[0])
        except Exception:
            pass
    
        prediction_value = sess.run(prediction,feed_dict={xs:x_data})
        lines = ax.plot(x_data,prediction_value,'r-',lw=5)
        #图像暂停
        plt.pause(0.1)

增加了对tensorboard的使用，包括graphs，scaler，histogram

import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt

#添加层的函数，层有输入参数，输出参数，偏置，还有激活函数
def add_layer(inputs,in_size,out_size,n_layer,activation_function=None):
    layer_name = 'layer%s'%n_layer
    #随机初始化参数和偏置
    with tf.name_scope('layer'):
        with tf.name_scope('weights'):
            Weights = tf.Variable(tf.random_normal([in_size,out_size]))
            tf.summary.histogram(layer_name+'/weights',Weights)
        with tf.name_scope('biase'):
            biases = tf.Variable(tf.zeros([1,out_size])+0.1)
            tf.summary.histogram(layer_name+'/biases',biases)
        #计算（W*a+b）
        with tf.name_scope('Wx_plus_b'):
            Wx_plus_b = tf.matmul(inputs,Weights)+biases
        #如果有激励函数就A（W*a+b）
        if activation_function is None:
            outputs = Wx_plus_b
        else:
            outputs = activation_function(Wx_plus_b)
        tf.summary.histogram(layer_name+'/outputs',outputs)
        return outputs
#生成训练数据
x_data = np.linspace(-1,1,300)[:,np.newaxis]
noise = np.random.normal(0,0.05,x_data.shape)
y_data = np.square(x_data)-0.5 + noise

#定义输入输出
with tf.name_scope('inputs'):
    xs=tf.placeholder(tf.float32,[None,1],name='x_input')
    ys=tf.placeholder(tf.float32,[None,1],name='y_input')

#生成l1层，1输入，19输出，relu激活
l1 = add_layer(xs,1,10,n_layer=1,activation_function=tf.nn.relu)
#生成prediction层，10输入，1输出，输出直接进行预测
prediction = add_layer(l1,10,1,n_layer=1,activation_function=None)

#定义loss函数，均方根误差
with tf.name_scope('loss'):
    loss = tf.reduce_mean(tf.reduce_sum(tf.square(ys-prediction),
                                    reduction_indices=[1]))
    tf.summary.scalar('loss',loss)
#定义训练方式：梯度下降法，误差最小化
with tf.name_scope('train'):
    train_step = tf.train.GradientDescentOptimizer(0.1).minimize(loss)
#初始化变量
init = tf.initialize_all_variables()
#建立会话
sess = tf.Session()
merged = tf.summary.merge_all()
writter = tf.summary.FileWriter("logs/",sess.graph)
#初始化
sess.run(init)

for i in range(1000):
    sess.run(train_step,feed_dict={xs:x_data,ys:y_data})
    if i%50 == 0:
        result = sess.run(merged,feed_dict={xs:x_data,ys:y_data})
        writter.add_summary(result,i)
        

reference：

https://www.bilibili.com/video/av16001891/?p=16