用零扩展矩阵的numpy数组 [英] Expanding a numpy array of matrices with zeros

问题描述

用零填充矩阵数组的最有效方法是什么?

What's the most efficient way to pad an array of matrices with zeros?

示例:

# Lets construct an array of 2 matrices from 3 arrays of vectors
import numpy as np

A = np.array([[0,1,2],[3,4,5]])       # 2 vectors
B = np.array([[6,7,8],[9,10,11]])     # 2 vectors
C = np.array([[12,13,14],[15,16,17]]) # 2 vectors
M = np.dstack((A,B,C))
'''
# Result: array([[[ 0,  6, 12],
                  [ 1,  7, 13],
                  [ 2,  8, 14]],

                 [[ 3,  9, 15],
                  [ 4, 10, 16],
                  [ 5, 11, 17]]]) #
'''

我想向数组中的每个矩阵元素添加一列和/或一行零，例如:

I want to add a column and/or a row of zeros to every matrix element in the array such as:

'''
# Result: array([[[ 0,  6, 12, 0],
              [ 1,  7, 13, 0],
              [ 2,  8, 14, 0],
              [ 0,  0,  0, 0]],

             [[ 3,  9, 15, 0],
              [ 4, 10, 16, 0],
              [ 5, 11, 17, 0]
              [ 0,  0,  0, 0]]]) #
'''

推荐答案

np.pad可以工作，但是在这种情况下，它是过大的.我们可以直接使用:

np.pad will work, but for this case it is overkill. We can do it directly with:

3d数组示例(不同的尺寸使更改更明显)

A sample 3d array (different dimensions make changes more obvious)

In [408]: M=np.arange(2*3*4).reshape((2,3,4))    
In [409]: M
Out[409]: 
array([[[ 0,  1,  2,  3],
        [ 4,  5,  6,  7],
        [ 8,  9, 10, 11]],

       [[12, 13, 14, 15],
        [16, 17, 18, 19],
        [20, 21, 22, 23]]])

具有所需目标形状的空白数组

A blank array of the desired target shape

In [410]: M1=np.zeros((2,4,5),M.dtype)

将M中的值复制到正确的切片范围内的目标.

Copy values from M to the target in the right slice range.

In [411]: M1[:,:-1,:-1]=M
In [412]: M1
Out[412]: 
array([[[ 0,  1,  2,  3,  0],
        [ 4,  5,  6,  7,  0],
        [ 8,  9, 10, 11,  0],
        [ 0,  0,  0,  0,  0]],

       [[12, 13, 14, 15,  0],
        [16, 17, 18, 19,  0],
        [20, 21, 22, 23,  0],
        [ 0,  0,  0,  0,  0]]])

需要这样的副本.无法扩展M本身的大小. pad也返回了一个新的数组，并执行了此分配和复制的常规版本.因此，效率问题不多.

A copy like this is required. There's no way of expanding the size of M itself. pad returns a new array as well, having performed a general version of this allocate and copy. So there isn't much of an efficiency issue.

您还可以在正确的维度中连接(或附加")0行或一列.但是我已经说明了一步.

You could also concatenate (or 'append') a 0 row or column in the right dimensions. But what I've illustrated does it in one step.

这篇关于用零扩展矩阵的numpy数组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！