tensorflow学习

import tensorflow as tf

labels = [[0,0,1],[0,1,0]]
logits = [[2,  0.5,6],
          [0.1,0,  3]]
logits_scaled = tf.nn.softmax(logits)
logits_scaled2 = tf.nn.softmax(logits_scaled)

result1 = tf.nn.softmax_cross_entropy_with_logits(labels=labels, logits=logits)
result2 = tf.nn.softmax_cross_entropy_with_logits(labels=labels, logits=logits_scaled)
result3 = -tf.reduce_sum(labels*tf.log(logits_scaled),1)

with tf.Session() as sess:
    print ("scaled=",sess.run(logits_scaled))    
    print ("scaled2=",sess.run(logits_scaled2)) #经过第二次的softmax后，分布概率会有变化
    print ("rel1=",sess.run(result1),"\n")#正确的方式
    print ("rel2=",sess.run(result2),"\n")#如果将softmax变换完的值放进去会，就相当于算第二次softmax的loss，所以会出错
    print ("rel3=",sess.run(result3))

('scaled=', array([[0.01791432, 0.00399722, 0.97808844],
[0.04980332, 0.04506391, 0.90513283]], dtype=float32))
('scaled2=', array([[0.21747023, 0.21446465, 0.56806517],
[0.2300214 , 0.22893383, 0.5410447 ]], dtype=float32))
('rel1=', array([0.02215516, 3.0996735 ], dtype=float32), '\n')
('rel2=', array([0.56551915, 1.4743223 ], dtype=float32), '\n')
('rel3=', array([0.02215518, 3.0996735 ], dtype=float32))

softmax的计算：

（e^2/(e2+e^0.5+e6),e^0.5/(e2+e^0.5+e6),e^6/(e2+e^0.5+e6)） = array([[0.01791432, 0.00399722, 0.97808844],

可是按照这个公式计算交叉熵的时候发现和结果不一样，看到result3 = -tf.reduce_sum(labels*tf.log(logits_scaled),1)，那就是直接log一下计算的？

2.tf.reduce_sum