tf.nn.embedding_lookup函数

# tensor是输入张量
# id就是张量对那个的索引
tf.nn.embedding_lookup(tensor, id)

该函数的用法主要是选取一个张量里面索引对应的元素，例如：

# 示例
import tensorflow as tf
import numpy as np

c = np.random.random([10, 1])
b = tf.nn.embedding_lookup(c, [1, 3])

with tf.Session() as sess:
    sess.run(tf.initialize_all_variables())
    print("b\n" + str(sess.run(b)))
    print("c\n" + str(c))

# 输出
b
[[0.16904814]
 [0.45015154]]
c
[[0.40655617]
 [0.16904814]
 [0.61787992]
 [0.45015154]
 [0.01681443]
 [0.46908087]
 [0.84033554]
 [0.92226145]
 [0.11049313]
 [0.0526304 ]]

参考自https://blog.youkuaiyun.com/UESTC_C2_403/article/details/72779417