[英]How in numpy get elements of matrix between two indices arrays?
Let's say I have a matrix: 假设我有一个矩阵:
>> a = np.arange(25).reshape(5, 5)`
>> a
[[ 0 1 2 3 4]
[ 5 6 7 8 9]
[10 11 12 13 14]
[15 16 17 18 19]
[20 21 22 23 24]]
and two vectors of indices that define a span of matrix elements that I want to extract: 和两个索引向量,用于定义我想要提取的矩阵元素的范围:
>> indices1 = np.array([0, 1, 1, 0, 0])
>> indices2 = np.array([2, 3, 3, 2, 2])
As you can see, difference between each corresponding index is equal to 2. 如您所见,每个相应索引之间的差异等于2。
I would like to do sth like this extract a part of the matrix: 我想做这样的提取矩阵的一部分:
>> submatrix = a[indices1:indices2, :]
so that the result would be 2x5 matrix: 这样结果就是2x5矩阵:
>> submatrix
[[ 0 6 7 3 4],
[ 5 11 12 8 9]]
For all I know, numpy allows to provide indices as a boundaries, but does not allow to provide arrays, only integers, eg a[0:2]
. 据我所知,numpy允许提供索引作为边界,但不允许提供数组,只允许整数,例如
a[0:2]
。
Note what I want to subtract is not a submatrix: 注意我想要减去的不是子矩阵:
Do you know of some other way of indexing a numpy matrix so that it is possible to provide arrays defining spans? 您是否知道索引numpy矩阵的其他方法,以便可以提供定义跨度的数组? For now I managed to do it only with for loops.
现在我设法只使用for循环。
For reference, the most obvious loop (still took several experimental steps): 作为参考,最明显的循环(仍然需要几个实验步骤):
In [87]: np.concatenate([a[i:j,n] for n,(i,j) in enumerate(zip(indices1,indices2))], ).reshape(-1,2).T
Out[87]:
array([[ 0, 6, 7, 3, 4],
[ 5, 11, 12, 8, 9]])
Broadcasted indices taking advantage of the constant length: 广播指数利用恒定长度:
In [88]: indices1+np.arange(2)[:,None]
Out[88]:
array([[0, 1, 1, 0, 0],
[1, 2, 2, 1, 1]])
In [89]: a[indices1+np.arange(2)[:,None],np.arange(5)]
Out[89]:
array([[ 0, 6, 7, 3, 4],
[ 5, 11, 12, 8, 9]])
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.