时间戳

首页 > 解决方案 > Softmax 不会导致 Python 实现中的概率分布

python - Softmax 不会导致 Python 实现中的概率分布

问题描述

我有一个简单的 softmax 实现：

softmax = np.exp(x) / np.sum(np.exp(x), axis=0)

对于 x 在此处设置为数组：https ://justpaste.it/6wis7

您可以将其加载为：

 import numpy as np

 x = np.as (just copy and paste the content (starting from array))

我得到：

softmax.mean(axis=0).shape 
(100,) # now all elements must be 1.0 here, since its a probability

softmax.mean(axis=0) # all elements are not 1

array([0.05263158, 0.05263158, 0.05263158, 0.05263158, 0.05263158,
       0.05263158, 0.05263158, 0.05263158, 0.05263158, 0.05263158,
       0.05263158, 0.05263158, 0.05263158, 0.05263158, 0.05263158,
       0.05263158, 0.05263158, 0.05263158, 0.05263158, 0.05263158,
       0.05263158, 0.05263158, 0.05263158, 0.05263158, 0.05263158,
       0.05263158, 0.05263158, 0.05263158, 0.05263158, 0.05263158,
       0.05263158, 0.05263158, 0.05263158, 0.05263158, 0.05263158,
       0.05263158, 0.05263158, 0.05263158, 0.05263158, 0.05263158,
       0.05263158, 0.05263158, 0.05263158, 0.05263158, 0.05263158,
       0.05263158, 0.05263158, 0.05263158, 0.05263158, 0.05263158,
       0.05263158, 0.05263158, 0.05263158, 0.05263158, 0.05263158,
       0.05263158, 0.05263158, 0.05263158, 0.05263158, 0.05263158,
       0.05263158, 0.05263158, 0.05263158, 0.05263158, 0.05263158,
       0.05263158, 0.05263158, 0.05263158, 0.05263158, 0.05263158,
       0.05263158, 0.05263158, 0.05263158, 0.05263158, 0.05263158,
       0.05263158, 0.05263158, 0.05263158, 0.05263158, 0.05263158,
       0.05263158, 0.05263158, 0.05263158, 0.05263158, 0.05263158,
       0.05263158, 0.05263158, 0.05263158, 0.05263158, 0.05263158,
       0.05263158, 0.05263158, 0.05263158, 0.05263158, 0.05263158,
       0.05263158, 0.05263158, 0.05263158, 0.05263158, 0.05263158])

为什么这个实现是错误的？如何解决？

标签： pythonpython-3.xmachine-learningdeep-learningsoftmax

解决方案

在我看来很好：

import numpy as np

def softmax(x):
    return np.exp(x) / np.sum(np.exp(x), axis=0)

logits = softmax(np.random.rand(4))

print(logits)

softmax 动作的所有元素的总和应该等于 1。

对于分类任务，通常采用具有最高值np.argmax()（

class_index = np.argmax(logits)  # Assuming logits is the output of a trained model

print('Most likely class: %d' % class_index)

正如 JosepJoestar 在评论中指出的那样，softmax 函数的定义可以在这里找到。

推荐阅读