tf.contrib.lookup.index_to_string_table_from_tensor和tf.nn.top_k()

最新推荐文章于 2020-12-10 12:15:30 发布

teleological_life

最新推荐文章于 2020-12-10 12:15:30 发布

阅读量1.9k

点赞数

CC 4.0 BY-SA版权

文章标签： Tensorflow

本文链接：https://blog.youkuaiyun.com/qq_37293427/article/details/82840279

本文详细解析了TensorFlow中tf.contrib.lookup.index_to_string_table_from_tensor函数的使用方法，包括一维和二维情况的示例，以及tf.nn.top_k函数的应用，通过实例展示了如何将索引转换为字符串，并获取概率最大的类别。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

tf.contrib.lookup.index_to_string_table_from_tensor(
    mapping,
    default_value='UNK',
    name=None
)

Returns a lookup table that maps a Tensor of indices into strings.

1.一维情况

import tensorflow as tf
sess=tf.Session()
mapping_string = tf.constant(["emerson", "lake", "palmer"])
indices1 = tf.constant([0,1,2], tf.int64)
indices2 = tf.constant([0,2], tf.int64)
indices3 = tf.constant([1], tf.int64)
table = tf.contrib.lookup.index_to_string_table_from_tensor(
    mapping_string)
values1 = table.lookup(indices1)
values2 = table.lookup(indices2)
values3 = table.lookup(indices3)
sess.run(tf.tables_initializer())

print(sess.run(mapping_string))

print(sess.run(indices1))
print(sess.run(values1))
print(sess.run(indices2))
print(sess.run(values2))
print(sess.run(indices3))
print(sess.run(values3))

运行结果:
[b’emerson’ b’lake’ b’palmer’]
[0 1 2]
[b’emerson’ b’lake’ b’palmer’]
[0 2]
[b’emerson’ b’palmer’]
[1]
[b’lake’]

2.二维情况

import tensorflow as tf
sess=tf.Session()
mapping_string = tf.constant(["emerson", "lake", "palmer"])
indices1 = tf.constant([[0,1],[1,2],[2,3]], tf.int64)

table = tf.contrib.lookup.index_to_string_table_from_tensor(
    mapping_string)
values1 = table.lookup(indices1)

sess.run(tf.tables_initializer())

print(sess.run(mapping_string))
print(sess.run(indices1))
print(sess.run(values1))

运行结果:
[b’emerson’ b’lake’ b’palmer’]
[[0 1]
[1 2]
[2 3]]
[[b’emerson’ b’lake’]
[b’lake’ b’palmer’]
[b’palmer’ b’UNK’]]

总结:按照张量indices的格式,返回indices中索引的值,格式不变.

3.tf.nn.top_k()
以下以一个5分类为例,一行表示一个样本属于5类的概率.按下面代码转化完了后,输出的每行第一个数字就代表该行样本所属的分类.

import tensorflow as tf
import numpy as np
sess=tf.Session()
y=np.array([[0.10,0.20,0.40,0.25,0.05],
            [0.45,0.15,0.10,0.05,0.25],
            [0.15,0.65,0.05,0.10,0.05]])

values,indices=tf.nn.top_k(y,5)
table = tf.contrib.lookup.index_to_string_table_from_tensor(
    tf.constant([str(i) for i in range(5)]))
prediction_classes = table.lookup(tf.to_int64(indices))
sess.run(tf.tables_initializer())
print(sess.run(values))
print(sess.run(indices))
print(sess.run(prediction_classes))

运行结果:
[[0.4 0.25 0.2 0.1 0.05]
[0.45 0.25 0.15 0.1 0.05]
[0.65 0.15 0.1 0.05 0.05]]
[[2 3 1 0 4]
[0 4 1 2 3]
[1 0 3 2 4]]
[[b’2’ b’3’ b’1’ b’0’ b’4’]
[b’0’ b’4’ b’1’ b’2’ b’3’]
[b’1’ b’0’ b’3’ b’2’ b’4’]]
总结: 返回的是每行由大到小排序的结果以及对应的它们在原来数组中的行索引.例如:第一行最大值为0.4,它在原数组中行索引是2(该行第三个数字).