仅在多维数组中“垂直”打乱特定列

Question

我有一个 6 列的多维数组，例如：

[59, '591', '592', '593', '594', 1582823720],
[9, '91', '92', '93', '94', 1582823745],
[7, '71', '72', '73', '74', 1582823745],
[61, '611', '612', '613', '614', 1582823752],
[54, '541', '542', '543', '544', 1582823717],
[24, '241', '242', '243', '244', 1582823706]

是否有一种简单的方法可以“垂直”地仅对特定列进行洗牌，同时保持其他列内容不变？

例如上面的例子，假设我只需要“垂直”打乱第 2-5 列，而将第 1 列和第 6 列保持原样，所以结果将是：

[59, '541', '242', '243', '74', 1582823720],
[9, '591', '542', '593', '94', 1582823745],
[7, '241', '612', '543', '614', 1582823745],
[61, '611', '92', '73', '544', 1582823752],
[54, '71', '72', '613', '594', 1582823717],
[24, '91', '592', '93', '244', 1582823706]

我是 Python 新手，也许有一个简单的内置解决方案或某个模块可以做到这一点？

我遇到了numpy 库，它使用random.shuffle() 函数可以“垂直”调整整个数组行变得轻而易举，也许有一个可以只调整特定列？

Answer 1

你可以用 numpy shuffle 函数来做到这一点

x=np.array(yourlist)    
np.random.shuffle(x[:,1:5])

对于水平随机播放，您可以使用转置

np.random.shuffle(x.T[:,1:5])

垂直洗牌示例

x = np.arange(36).reshape(6,6)
x
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],
       [24, 25, 26, 27, 28, 29],
       [30, 31, 32, 33, 34, 35]])
np.random.shuffle(x[:,1:5])
x
array([[ 0,  7,  8,  9, 10,  5],
       [ 6,  1,  2,  3,  4, 11],
       [12, 19, 20, 21, 22, 17],
       [18, 25, 26, 27, 28, 23],
       [24, 13, 14, 15, 16, 29],
       [30, 31, 32, 33, 34, 35]])

Answer 2

这是一个使用 numpy 的代码。

data = [[59, '541', '242', '243', '74', 1582823720],
    [9, '591', '542', '593', '94', 1582823745],
    [7, '241', '612', '543', '614', 1582823745],
    [61, '611', '92', '73', '544', 1582823752],
    [54, '71', '72', '613', '594', 1582823717],
    [24, '91', '592', '93', '244', 1582823706]
]


import numpy as np
import random 


data_numpy = np.array(data)


def shuffle_column(matrix, col_index_to_shuffle):
  """
  """
  current_data = matrix[:, col_index_to_shuffle]
  random.shuffle(current_data)
  matrix[:, col_index_to_shuffle] = current_data
  return matrix


shuffled_matrix = shuffle_column(data_numpy, 2)
shuffled_matrix

array([['59', '541', '242', '243', '74', '1582823720'],
       ['9', '591', '92', '593', '94', '1582823745'],
       ['7', '241', '592', '543', '614', '1582823745'],
       ['61', '611', '612', '73', '544', '1582823752'],
       ['54', '71', '72', '613', '594', '1582823717'],
       ['24', '91', '542', '93', '244', '1582823706']], dtype='<U21')

Answer 3

我不确定这是否存在于其他库中，尽管我相信这样的功能应该存在。但是，我不需要numpy 来做到这一点：

• 转置数组。
• 随机排列您需要的子数组。
• 将数组转回。

洗牌第四列的代码示例是：

import random
# I am using pprint to beautify the output on the terminal
from pprint import pprint
arr = [[59, '591', '592', '593', '594', 1582823720],
       [9, '91', '92', '93', '94', 1582823745],
       [7, '71', '72', '73', '74', 1582823745],
       [61, '611', '612', '613', '614', 1582823752],
       [54, '541', '542', '543', '544', 1582823717],
       [24, '241', '242', '243', '244', 1582823706]
      ]
t_arr = [*zip(*arr)]
# I am converting array elements to lists as the zip() function produce tuples instead of lists.
t_arr = [list(sub_arr) for sub_arr in t_arr]
random.shuffle(t_arr[3])
arr_b = [*zip(*t_arr)]
# Again, converting back to lists
arr_b = [list(sub_arr) for sub_arr in arr_b]
# printing out the results :)
pprint(arr_b)

这是输出：

[[59, '591', '592', '73', '594', 1582823720],
 [9, '91', '92', '243', '94', 1582823745],
 [7, '71', '72', '543', '74', 1582823745],
 [61, '611', '612', '93', '614', 1582823752],
 [54, '541', '542', '613', '544', 1582823717],
 [24, '241', '242', '593', '244', 1582823706]]

Answer 4

numpy shuffle 可以就地打乱子数组。

如果你想让 4 列保持它们的水平一致性，就这样做

data = np.array(data)
np.random.shuffle(data[1:5])

前面有np.random.seed(0)，它给出了

array([['59', '591', '592', '593', '594', '1582823720'],
       ['61', '611', '612', '613', '614', '1582823752'],
       ['54', '541', '542', '543', '544', '1582823717'],
       ['7', '71', '72', '73', '74', '1582823745'],
       ['9', '91', '92', '93', '94', '1582823745'],
       ['24', '241', '242', '243', '244', '1582823706']], dtype='<U11')

如果您希望列单独打乱：

data = np.array(data)
tdata = np.transpose(tdata)
for i in range(1,5): np.shuffle(tdata[i])
data = np.transpose(tdata)

前面有np.random.seed(0)，它给出了

array([['59', '241', '92', '613', '244', '1582823720'],
       ['9', '71', '612', '243', '74', '1582823745'],
       ['7', '91', '542', '93', '614', '1582823745'],
       ['61', '611', '592', '73', '544', '1582823752'],
       ['54', '591', '72', '543', '94', '1582823717'],
       ['24', '541', '242', '593', '594', '1582823706']], dtype='<U11')