如何从多个标签中选择一个最常见的标签？使用 python

Question

我有加速度（只有 x 轴）信号，对于每个样本，我都有一个标签（编码为一个热代码）。

-我的加速度信号是这样的：

0     0.294132
1    -0.499688
2    -2.178345
3    -2.797763
4    -2.167961
5    -1.734857
6    -1.463939
7    -0.945008
8     0.182038
9     0.948368
10    1.675603

    ...   
95    0.793209
96   -0.023213
97    0.284969
98   -0.170126
99   -2.005471
Name: X-Axis, Length: 100, dtype: float64

-这是每个标签的样子：

array([[0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 1, 0, 0, 0],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 1, 0, 0, 0, 0],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1],
   [0, 0, 0, 0, 0, 1]], dtype=int8)

如何 select 每 100 个样本一个标签？换句话说，我如何将这个标签数组从 (100,6) 压缩到 (1,6)，使得出现最多的标签（概率最高的是 selected）

Answer 1

假设 a 是您在上面发布的数组，您可以在维度 0 上对数组 a 求和，找到所得总和的最大值并提取其位置：

import tensorflow as tf

labels = tf.constant(a, dtype=tf.int32)

# count number of labels for every position
count = tf.reduce_sum(labels, axis=[0])
# count has shape (6)

# get the label with the higher count value
top_label = tf.argmax(count)

# One hot encode the label to get the same representation
one_hot_top_label = tf.one_hot(top_label, count.shape[0], dtype=tf.int32)


sess = tf.InteractiveSession()
print(sess.run([count, top_label, one_hot_top_label]))

你得到：

[array([ 0,  1,  1,  0,  0, 98], dtype=int32), 
5,
array([0, 0, 0, 0, 0, 1], dtype=int32)]

如何从多个标签中选择一个最常见的标签？使用 python

How to choose a single most common label from multiple labels? using python

python

tensorflow

convolutional-neural-network