我更喜欢下一个提示: 将矩阵命名为matrix_a并使用column_number，例如:

import numpy as np
matrix_a = np.array([[1,2,3,4],[5,6,7,8],[9,10,11,12]])
column_number=2

# you can get the row from transposed matrix - it will be a column:
col=matrix_a.transpose()[column_number]

假设我们有nxm矩阵(n行m列)5行4列

matrix = [[1,2,3,4],[5,6,7,8],[9,10,11,12],[13,14,15,16],[17,18,19,20]]

要在python中提取列，我们可以像这样使用列表推导式

[ [row[i] for row in matrix] for in range(4) ]

你可以用矩阵的列数来替换4。 结果是

,10,14,18,5,9,13,17 [[1], [2], [3,7,11,15,19], [4,8,12,16,20]]

所有列从一个矩阵到一个新的列表:

N = len(matrix) 
column_list = [ [matrix[row][column] for row in range(N)] for column in range(N) ]

>>> import numpy as np
>>> A = np.array([[1,2,3,4],[5,6,7,8]])

>>> A
array([[1, 2, 3, 4],
    [5, 6, 7, 8]])

>>> A[:,2] # returns the third columm
array([3, 7])

参见:"numpy。“Arange”和“重塑”来分配内存

示例:(用矩阵(3x4)的形状分配数组)

nrows = 3
ncols = 4
my_array = numpy.arange(nrows*ncols, dtype='double')
my_array = my_array.reshape(nrows, ncols)

如果你在Python中有一个二维数组(不是numpy)，你可以像这样提取所有的列，

data = [
['a', 1, 2], 
['b', 3, 4], 
['c', 5, 6]
]

columns = list(zip(*data))

print("column[0] = {}".format(columns[0]))
print("column[1] = {}".format(columns[1]))
print("column[2] = {}".format(columns[2]))

执行这段代码会得到，

>>> print("column[0] = {}".format(columns[0]))
column[0] = ('a', 'b', 'c')

>>> print("column[1] = {}".format(columns[1]))
column[1] = (1, 3, 5)

>>> print("column[2] = {}".format(columns[2]))
column[2] = (2, 4, 6)

如果你喜欢map-reduce风格的python, itemgetter操作符也会有帮助，而不是列表推导式，为了一点变化!

# tested in 2.4
from operator import itemgetter
def column(matrix,i):
    f = itemgetter(i)
    return map(f,matrix)

M = [range(x,x+5) for x in range(10)]
assert column(M,1) == range(1,11)