展开numpy矩阵 [英] Expand numpy matrix

问题描述

我正在尝试以某种方式扩展numpy矩阵，通常如下所示:

I am trying to somehow expand numpy matrices, which typically look like:

import numpy as np

mtx = np.matrix([[['a','b','c'], ['x'], 3], [['d','e','f'], ['y'], 2],
    [['g','h','i'], ['z'], 1]])
mtx
# matrix([[['a', 'b', 'c'], ['x'], 3],
#         [['d', 'e', 'f'], ['y'], 2],
#         [['g', 'h', 'i'], ['z'], 1]], dtype=object)

最后一列包含生成的矩阵的实例数，然后应如下所示:

The last column contains number of instances of the resulting matrix, which then should look like this:

# matrix([[['a', 'b', 'c'], ['x']],
#         [['a', 'b', 'c'], ['x']],
#         [['a', 'b', 'c'], ['x']],
#         [['d', 'e', 'f'], ['y']],
#         [['d', 'e', 'f'], ['y']],
#         [['g', 'h', 'i'], ['z']]], dtype=object)

所以，第一行三遍，第二行两遍，等等.

So, three times 1st row, two times 2nd etc.

我想知道最快和/或最优雅的python-way是什么吗?

I wonder what would be the fastest and/or the most elegant python-way?

许多tnx！下午

推荐答案

您可以使用 np.repeat 重复每行mtx[:,:2]的前两列，由第三列arr[:,2]的相应行给出的次数:

You could use np.repeat to repeat the first two columns of each row mtx[:,:2] the number of times given by the corresponding row of the third column arr[:,2]:

>>> arr = np.asarray(mtx)
>>> np.repeat(arr[:,:2], arr[:,2].astype(int), axis=0)
array([[['a', 'b', 'c'], ['x']],
       [['a', 'b', 'c'], ['x']],
       [['a', 'b', 'c'], ['x']],
       [['d', 'e', 'f'], ['y']],
       [['d', 'e', 'f'], ['y']],
       [['g', 'h', 'i'], ['z']]], dtype=object)

第三列需要首先转换为整数值(例如，使用astype(int)).我还发现必须将mtx视为array才能起作用:您可以使用np.matrix轻松地将其再次转换为matrix对象.

The third column needs to be cast to integer values first (e.g. using astype(int)). I also found it necessary to treat mtx as an array for this to work: you can easily turn it back into a matrix object again with np.matrix.

这篇关于展开numpy矩阵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！