: 和 , 在 numpy 中的区别 [英] Difference between : and , in numpy

问题描述

一些资源提到在numpy的数组切片中，array[2,:,1]的结果与array[2][:][1] ，但在这种情况下我没有得到相同的:

Some resources have mentioned that in numpy's array slicing, array[2,:,1] results in the same as array[2][:][1] , but I do not get the same ones in this case:

array3d = np.array([[[1, 2], [3, 4]],[[5, 6], [7, 8]], [[9, 10], [11, 12]]])
array3d[2,:,1]
out: array([10, 12])

和:

array3d[2][:][1]
out: array([11, 12])

有什么区别?

推荐答案

一些资源是错误的！

In [1]: array3d = np.array([[[1, 2], [3, 4]],[[5, 6], [7, 8]], [[9, 10], [11, 12
   ...: ]]])
In [2]: array3d
Out[2]: 
array([[[ 1,  2],
        [ 3,  4]],

       [[ 5,  6],
        [ 7,  8]],

       [[ 9, 10],
        [11, 12]]])

当索引都是标量时，这种分解有效:

When the indices are all scalar this kind of decomposition works:

In [3]: array3d[2,0,1]
Out[3]: 10
In [4]: array3d[2][0][1]
Out[4]: 10

一个索引减少了维度，选择了一个平面":

One index reduces the dimension, picking one 'plane':

In [5]: array3d[2]
Out[5]: 
array([[ 9, 10],
       [11, 12]])

[:] 上什么都不做 - 它本身不是占位符.在多维索引中，它是一个切片——该维度中的整个事物.我们在列表中看到了相同的行为.alist[2] 返回一个元素，alist[:] 返回整个列表的副本.

[:] on that does nothing - it is not a place holder by itself. Within the multidimensional index it is a slice - the whole thing in that dimension. We see the same behavior with lists. alist[2] returns an element, alist[:] returns a copy of the whole list.

In [6]: array3d[2][:]
Out[6]: 
array([[ 9, 10],
       [11, 12]])

记住，numpy 是一个 python 包.Python 语法仍然适用于所有级别.x[a][b][c] 按顺序进行 3 次索引操作，链接"它们.x[a,b,c] 是一种索引操作，将 的元组传递给 x.这是解释该元组的 numpy 代码.

Remember, numpy is a python package. Python syntax still applies at all levels. x[a][b][c] does 3 indexing operations in sequence, 'chaining' them. x[a,b,c] is one indexing operation, passing a tuple of to x. It's numpy code that interprets that tuple.

我们必须在剩余维度上使用多维索引:

We have to use a multidimensional index on the remaining dimensions:

In [7]: array3d[2][:,1]
Out[7]: array([10, 12])
In [8]: array3d[2,:,1]
Out[8]: array([10, 12])

解释器实际上做了:

In [9]: array3d.__getitem__((2,slice(None),1))
Out[9]: array([10, 12])

In [11]: array3d.__getitem__(2).__getitem__((slice(None),1))
Out[11]: array([10, 12])

这篇关于: 和 , 在 numpy 中的区别的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！