【機器學習】搭建神經網路筆記
阿新 • • 發佈:2018-12-31
一、簡單寫一個迴歸方程
import tensorflow as tf import numpy as np #creat data x_data = np.random.rand(100).astype(np.float32)#在x中生成隨機數,隨機數以np的float32型別展示 y_data= x_data * 0.1 + 0.3 #基本的函式 # create tensorflow structure start# Weights= tf.Variable(tf.random_uniform([1] , -1.0 , 1.0))#初始化Weights(權重)的張量,均勻分佈 biases = tf.Variable(tf.zeros([1]))#初始化biases(偏移量)張量,一維的資料 y = Weights * x_data + biases #依據的Weight和biases兩個建立一個模型 lost = tf.reduce_mean(tf.square(y - y_data))#lost的數值為求得的是(y-y.data)^2的平均值 optimizer = tf.train.GradientDescentOptimizer(0.5)#梯度下降優化器,範圍為0.5 train = optimizer.minimize(lost) init = tf.global_variables_initializer(); # create tensorflow structure end # sess = tf.Session()#建立訪問 sess.run(init) #執行 for step in range(201): sess.run(train) if(step % 20) == 0: print(step , sess.run(Weights), sess.run(biases))
二、tensorflow的會話機制:Session
#Session的兩種寫法 import tensorflow as tf martix1 = tf.constant([[3 , 3]]) martix2 = tf.constant([[2], [2]]) product = tf.matmul(martix1 , martix2) # #method 1 # sess = tf.Session()#Session記得要大寫 # result = sess.run(product) # print(result) # sess.close() #method 2 with tf.Session() as sess: result2 = sess.run(product) print(result2)
三、tensorflow的初始化機制:Variable
#Variable:建立變數 import tensorflow as tf state = tf.Variable(0 , name = 'counter') # print(state.name) one = tf.constant(1) new_value = tf.add(state , one) update = tf.assign(state , new_value)#assign:轉讓編制;將new_value賦值給state,return state #使用tf.global_variables_initializer()新增節點用於初始化所有的變數。 #在你構建完整個模型並在會話中載入模型後,執行這個節點。c231 init = tf.global_variables_initializer()#初始化模型 with tf.Session() as sess: sess.run(init) for _ in range(3): sess.run(update) print(sess.run(state))
四、placeholder
#placeholder:在執行的時候再去給我的值,而不是一開始就先賦值。
import tensorflow as tf
input1 = tf.placeholder(tf.float32)
input2 = tf.placeholder(tf.float32)
output = tf.multiply(input1 , input2)
with tf.Session() as sess:
print(sess.run(output , feed_dict = {input1:[7.] , input2:[2.]}))#將feed_dict的數值傳入output
五、搭建一個神經網路
#定義一個新增層
#建造神經網路
%matplotlib qt5
import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
#新增激勵函式
def add_layer(inputs, in_size, out_size, activation_function = None):
with tf.name_scope('layer'):
Weights = tf.Variable(tf.random_normal([in_size , out_size]) , name = 'W')
with tf.name_scope('biases'):
biases = tf.Variable(tf.zeros([1 , out_size]) + 0.1 , name = 'b')#初始化讓所有的數值都是0.1
with tf.name_scope('Wx_plus_b'):
Wx_plus_b = tf.matmul(inputs , Weights) + biases #矩陣的乘法,表示式
if activation_function is None: #沒有激勵的話直接輸出
outputs = Wx_plus_b
else:
outputs = activation_function(Wx_plus_b) #有激勵的話就做激勵
return outputs
#定義資料形式
x_data = np.linspace(-1 , 1 , 300)[: , np.newaxis]#300行有300個例子
noise = np.random.normal(0 , 0.05 , x_data.shape)#形成一些噪點
y_data = np.square(x_data) - 0.5 + noise
with tf.name_scope('inputs'):
xs = tf.placeholder(tf.float32 , [None , 1] , name = 'x_input')#傳進來的數值
ys = tf.placeholder(tf.float32 , [None , 1] , name = 'y_input')#傳進來的數值 ?????
l1 = add_layer(xs , 1 , 10 , activation_function = tf.nn.relu)#隱藏層,10個因子
prediction = add_layer(l1 , 10 , 1 , activation_function = None)#輸出層
with tf.name_scope('loss'):
loss = tf.reduce_mean(tf.reduce_sum(tf.square(ys - prediction) , reduction_indices = [1]))#求誤差
with tf.name_scope('train'):
train_step = tf.train.GradientDescentOptimizer(0.1).minimize(loss)#梯度下降演算法,學習效率0.1
init = tf.global_variables_initializer()
sess = tf.Session()
writer = tf.summary.FileWriter("logs/" , sess.graph)
sess.run(init)
fig = plt.figure()
ax = fig.add_subplot(1 , 1 , 1)
ax.scatter(x_data , y_data)
plt.ion()
plt.show()
#plt.ioff()
for i in range(1000):
sess.run(train_step , feed_dict = {xs:x_data , ys:y_data})
if i % 50 == 0:
try:
ax.lines.remove(lines[0])
except Exception:
pass
#print(sess.run(loss , feed_dict = {xs:x_data , ys:y_data}))
prediction_value = sess.run(prediction , feed_dict = {xs : x_data})
lines = ax.plot(x_data , prediction_value , 'r-' , lw = 5)
plt.pause(0.1)