loss = tf.reduce_sum(tf.where(tf.greater(y_, y), (y_ - y) * loss_less, (y - y_) * loss_more))

import tensorflow as tf
v1 = tf.constant([1.0, 2.0, 3.0, 4.0])
v2 = tf.constant([4.0, 3.0, 2.0, 1.0])
with tf.Session() as sess:
 print(sess.run(tf.greater(v1, v2)))
 print(sess.run(tf.where(tf.greater(v1, v2), v1, v2)))
 '''輸出結(jié)果為：
 [False False True True]
 [4. 3. 3. 4.]'''

import tensorflow as tf
from numpy.random import RandomState

#聲明wl、W2兩個(gè)變量，通過(guò)seed參數(shù)設(shè)定了隨機(jī)種子，這樣可以保證每次運(yùn)行得到的結(jié)果是一樣的
w = tf.Variable(tf.random_normal([2, 1], stddev=1, seed=1))

x = tf.placeholder(tf.float32, shape=(None, 2), name="x-input")
y_ = tf.placeholder(tf.float32, shape=(None, 1), name="y-input")

#定義神經(jīng)網(wǎng)絡(luò)結(jié)構(gòu)
y = tf.matmul(x, w)

#定義真實(shí)值與預(yù)測(cè)值之間的交叉熵?fù)p失函數(shù)，來(lái)刻畫真實(shí)值與預(yù)測(cè)值之間的差距
loss_less = 10
loss_more = 1
loss = tf.reduce_sum(tf.where(tf.greater(y_, y), (y_ - y) * loss_less, (y - y_) * loss_more))

#定義反向傳播算法的優(yōu)化方法
train_step = tf.train.AdamOptimizer(learning_rate=0.001).minimize(loss)

#設(shè)置隨機(jī)數(shù)種子
rdm = RandomState(seed=1)
#設(shè)置隨機(jī)數(shù)據(jù)集大小
dataset_size = 128
X = rdm.rand(dataset_size, 2)
'''設(shè)置回歸的正確值為兩個(gè)輸入的和加上一個(gè)隨機(jī)量。
之所以要加上一個(gè)隨機(jī)量是為了加入不可預(yù)測(cè)的噪音，否則不同損失函數(shù)的意義就不大了，因?yàn)椴煌瑩p失函數(shù)都會(huì)在能完全預(yù)測(cè)正確的時(shí)候最低。
一般來(lái)說(shuō)噪音為一個(gè)均值為0的小量，所以這里的噪音設(shè)置為-0.05——0.05的隨機(jī)數(shù)。'''
Y = [[x1 + x2 + rdm.rand()/10.0 -0.05] for x1,x2 in X]

#創(chuàng)建會(huì)話
with tf.Session() as sess:
 #初始化變量
 init_op = tf.global_variables_initializer()
 sess.run(init_op)
 
 print(sess.run(w))
 
 #設(shè)置batch訓(xùn)練數(shù)據(jù)的大小
 batch_size = 8
 #設(shè)置訓(xùn)練得輪數(shù)
 STEPS = 5000
 for i in range(STEPS):
  #每次選取batch_size個(gè)樣本進(jìn)行訓(xùn)練
  start = (i * batch_size) % dataset_size
  end = min(start + batch_size, dataset_size)

  #通過(guò)選取的樣本訓(xùn)練神經(jīng)網(wǎng)絡(luò)并更新參數(shù)
  sess.run(train_step, feed_dict={x:X[start:end], y_:Y[start:end]})
 
 print(sess.run(w))
 '''輸出結(jié)果為：
 [[-0.8113182]
 [ 1.4845988]]
 [[1.019347 ]
 [1.0428089]]'''

loss_less = 1
loss_more = 10

[[-0.8113182]
 [ 1.4845988]]
[[0.95561105]
 [0.98101896]]