Question

当我使用np.stack时，有时必须使用axis，就像axis = 1。我不明白轴的意义。例如，

c1 = np.ones((2, 3))
c2 = np.zeros((2, 3))
c = np.stack([c1, c2], axis = 1)

这显示了，

array([[[1., 1., 1.],
        [0., 0., 0.]],

       [[1., 1., 1.],
        [0., 0., 0.]]])

结果是什么规则？

Answer 1

轴表示尺寸。举一个简单的例子，考虑numpy.sum

import numpy as np
a=np.array([1,2,3],[2,3,1])
sum1=np.sum(a,axis=0)
sum2=np.sum(a,axis=1)
print sum1,sum2

这将给我sum1 = 12和sum2 = [3,5,4]

我的数组有两个维度/轴。第一个长度为2，第二个长度为3.因此，通过指定轴，您可以简单地告诉您的代码您希望在哪个维度上完成工作。

numpy.ndarray.ndim可以告诉你有多少轴

Answer 2

实际上sum1给出（3,5,4），sum2给出（6,6）但是你的评论让我了解了轴在Numpy中是如何工作的。谢谢。据我所知，axis = 0表示它将沿列添加，axis = 1表示它将沿行添加。

Answer 3

In your example, the arrays are 2d, and axis normally refers to one of those 2 dimensions.

In [441]: c1
Out[441]: 
array([[ 1.,  1.,  1.],
       [ 1.,  1.,  1.]])
In [442]: c1.sum(axis=0)
Out[442]: array([ 2.,  2.,  2.])
In [443]: c1.sum(axis=1)
Out[443]: array([ 3.,  3.])

Exactly what a function does with the axis parameter is up to the function itself. In the sum case is adds 'along' that axis, and returns a value that is 'missing' that axis. It's easier to see that action than it is to describe it.

The role of axis in concatenate is illustrated by:

In [452]: np.concatenate((c1,c2),axis=0).shape
Out[452]: (4, 3)
In [453]: np.concatenate((c1,c2),axis=1).shape
Out[453]: (2, 6)

stack adds a dimension. It's a relatively new API to concatenate, and works by first adding a dimension to each input array

In [448]: np.stack((c1,c2,c1,c2),axis=0).shape
Out[448]: (4, 2, 3)
In [449]: np.stack((c1,c2,c1,c2),axis=1).shape
Out[449]: (2, 4, 3)
In [450]: np.stack((c1,c2,c1,c2),axis=2).shape
Out[450]: (2, 3, 4)

什么是Python中的轴与Numpy模块？

3 个答案: