Tensorflow%20实战Google深度学习框架 4.2.2 自定义损失函数源代码

import os
import tab
import tensorflow as tf
from numpy.random import RandomState
print "hello tensorflow 4.1"
 
batch_size = 8
 
x = tf.placeholder(tf.float32,shape=(None,2),name='x-input')
y_ = tf.placeholder(tf.float32,shape=(None,1),name='y-input')
 
w1 = tf.Variable(tf.random_normal([2,1],stddev=1,seed=1))
#w2 = tf.Variable(tf.random_normal([3,1],stddev=1,seed=1))
y = tf.matmul(x,w1)
 
#a = tf.matmul(x,w1)
#y = tf.matmul(a,w2)
 
loss_less = 10
loss_more = 1
loss = tf.reduce_sum(tf.where(tf.greater(y,y_),(y-y_)*loss_more,(y_-y)*loss_less))
train_step = tf.train.AdamOptimizer(0.001).minimize(loss)
 
rdm = RandomState(1)
dataset_size = 128
X = rdm.rand(dataset_size,2)
Y = [[x1 + x2 +rdm.rand()/10.0-0.05] for (x1 ,x2 ) in X]
 
with tf.Session() as sess:
    init_op = tf.global_variables_initializer()
    sess.run(init_op)
    print sess.run(w1)
    STEPS = 5000
    for i in range(STEPS):
        start = (i * batch_size) % dataset_size
        end = min(start+batch_size,dataset_size)
        sess.run(train_step, feed_dict = {x: X[start:end], y_: Y[start:end]} )
        print sess.run(w1)
 
print "end "