tensorflow——tf.one_hot以及tf.sparse_to_dense函数

最新推荐文章于 2023-01-13 11:53:22 发布

原创最新推荐文章于 2023-01-13 11:53:22 发布 · 7.4k 阅读

0 ·

CC 4.0 BY-SA版权

本文通过实例详细介绍了使用 TensorFlow 中的 tf.one_hot 和 tf.sparse_to_dense 函数进行 one-hot 编码的方法。展示了如何将整数标签转换为 one-hot 向量，并比较了不同函数的应用场景。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

1、tf.one_hot函数

import numpy as np
import tensorflow as tf

SIZE=6
CLASS=10
label1=np.random.randint(0,10,size=SIZE) 

b = tf.one_hot(label1,CLASS,1,0)
with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    sess.run(b)
    print(sess.run(b))

输出结果：
产生的随机数：[7, 2, 9, 8, 4, 2]

[[ 0.  0.  0.  0.  0.  0.  0.  1.  0.  0.]
 [ 0.  0.  1.  0.  0.  0.  0.  0.  0.  0.]
 [ 0.  0.  0.  0.  0.  0.  0.  0.  0.  1.]
 [ 0.  0.  0.  0.  0.  0.  0.  0.  1.  0.]
 [ 0.  0.  1.  0.  0.  0.  0.  0.  0.  0.]
 [ 0.  0.  0.  0.  1.  0.  0.  0.  0.  0.]]

2、tf.sparse_to_dense函数

import tensorflow as tf   
import numpy as np

SIZE=6
CLASS=10
label=np.random.randint(0,10,size=SIZE) 
label=np.reshape(label,[SIZE,1])
index = np.reshape(np.arange(SIZE), [SIZE, 1])
#use a matrix  
concated = tf.concat([index, label], 1)  
onehot_labels = tf.sparse_to_dense(concated, [SIZE, CLASS], 1.0, 0.0)  

#use a vector  
concated2=tf.constant([1,3,4])  
onehot_labels2 = tf.sparse_to_dense(concated2, [ CLASS], 1.0, 0.0)

#use a scalar  
concated3=tf.constant(5)  
onehot_labels3 = tf.sparse_to_dense(concated3, [ CLASS], 1.0, 0.0)  

with tf.Session() as sess:  
    sess.run(tf.global_variables_initializer())
    result1=sess.run(onehot_labels)  
    result2 = sess.run(onehot_labels2)  
    result3 = sess.run(onehot_labels3)  
    print ("This is result1:")  
    print (result1)  
    print ("This is result2:")  
    print (result2)  
    print ("This is result3:")  
    print (result3)

输出结果：
产生的随机数：[7, 2, 9, 8, 4, 2]

This is result1:
[[ 0.  0.  0.  0.  0.  0.  0.  1.  0.  0.]
 [ 0.  0.  1.  0.  0.  0.  0.  0.  0.  0.]
 [ 0.  0.  0.  0.  0.  0.  0.  0.  0.  1.]
 [ 0.  0.  0.  0.  0.  0.  0.  0.  1.  0.]
 [ 0.  0.  1.  0.  0.  0.  0.  0.  0.  0.]
 [ 0.  0.  0.  0.  1.  0.  0.  0.  0.  0.]]
This is result2:
[ 0.  1.  0.  1.  1.  0.  0.  0.  0.  0.]
This is result3:
[ 0.  0.  0.  0.  0.  1.  0.  0.  0.  0.]