hypc's blog
0%

Numpy数组维度变换

Posted on Disqus:

reshape

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
>>> a = arange(24)
>>> a
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])
>>> b = a.reshape((2, 3, 4))
>>> b
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]]])
>>> c = b.reshape((3, 8))
>>> c
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]])

注意：reshape只是返回数组的一个视图（View），并没有分配内存保存结果。 也就是说，当改变其中某个元素的值时，原数组中的值也会改变。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
>>> b
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]]])
>>> b[0,0,0] = 100
>>> b
array([[[ 100,  1,  2,  3],
        [ 4,  5,  6,  7],
        [ 8,  9, 10, 11]],

       [[12, 13, 14, 15],
        [16, 17, 18, 19],
        [20, 21, 22, 23]]])
>>> a
array([ 100,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16,
       17, 18, 19, 20, 21, 22, 23])
>>> c
array([[ 100,  1,  2,  3,  4,  5,  6,  7],
       [ 8,  9, 10, 11, 12, 13, 14, 15],
       [16, 17, 18, 19, 20, 21, 22, 23]])

ravel

用来将多维数组展平。

1
2
3
4
5
6
7
8
9
10
11
12
13
>>> a = arange(24).reshape((2, 3, 4))
>>> a
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]]])
>>> b = a.ravel()
>>> b
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])

注意：ravel与reshape一样，也只是返回数组的一个视图（View），并没有分配内存保存结果。 也就是说，当改变其中某个元素的值时，原数组中的值也会改变。

flatten

这个函数同样用来将多维数组展平，但与ravel不同的是，这个函数会分配内存将结果保存起来。

1
2
3
4
5
6
7
8
9
10
11
12
13
>>> a = arange(24).reshape((2, 3, 4))
>>> a
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]]])
>>> b = a.flatten()
>>> b
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])

通过shape属性设置维度

通过直接给shape赋值来修改维度。

1
2
3
4
5
6
7
8
9
10
11
12
13
>>> a = arange(24).reshape((2, 3, 4))
>>> a
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]]])
>>> a.shape = (2, 12)
>>> a
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]])

resize

1
2
3
4
5
6
7
8
9
10
11
12
13
>>> a = arange(24).reshape((2, 3, 4))
>>> a
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]]])
>>> a.resize((2, 12))
>>> a
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]])

transpose

transpose用来转换矩阵。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
>>> a = arange(24).reshape((4, 6))
>>> a
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]])
>>> b = a.transpose()
>>> b
array([[ 0,  6, 12, 18],
       [ 1,  7, 13, 19],
       [ 2,  8, 14, 20],
       [ 3,  9, 15, 21],
       [ 4, 10, 16, 22],
       [ 5, 11, 17, 23]])

注意：同样，transpose也只是返回数组的一个视图（View），并没有分配内存保存结果。 也就是说，当改变其中某个元素的值时，原数组中的值也会改变。