np.broadcast_to使用举例

将数组广播到新的形状
举例说明np.broadcast函数的作用以及如何使用。

例1：将一行四列的数组广播到 四行四列：

import numpy as np
a = np.arange(4).reshape(1,4)
print(a)
print (np.broadcast_to(a,(4,4)))

输出如下：

[[0 1 2 3]]
[[0 1 2 3]
 [0 1 2 3]
 [0 1 2 3]
 [0 1 2 3]]

例2：如果将一行三列的数组广播到四行四列就会报错

"""报错"""
a = np.arange(3).reshape(1,3)
print(a)
print (np.broadcast_to(a,(4,4)))

提示

operands could not be broadcast together with remapped shapes [original->remapped]: (1,3) and requested shape (4,4)

例3：将一行三列的数组广播到四行三列

a = np.arange(3).reshape(1,3)
print(a)
print (np.broadcast_to(a,(4,3)))

成功运行

[[0 1 2]]
[[0 1 2]
 [0 1 2]
 [0 1 2]
 [0 1 2]]

例4：将三行一列的数组广播到四行三列，报错

"""报错"""
a = np.arange(3).reshape(3,1)
print(a)
print (np.broadcast_to(a,(4,3)))

报错如下：

operands could not be broadcast together with remapped shapes [original->remapped]: (3,1) and requested shape (4,3)

例5：将三行一列的数组广播到三行四列，成功

a = np.arange(3).reshape(3,1)
print(a)
print (np.broadcast_to(a,(3,4)))

输出如下：

[[0]
 [1]
 [2]]
[[0 0 0 0]
 [1 1 1 1]
 [2 2 2 2]]

例子6：将一行2列广播到（3, 4，2）

如果广播导(3， 4， 1）就会报错

import numpy as np
a = [1, 2]
# shape = (3, 4, 5)3, 4, 5会报错
b  = np.broadcast_to(a ,(3, 4, 2))
print(b)
print(b.shape)

输出如下

[[[1 2]
  [1 2]
  [1 2]
  [1 2]]

 [[1 2]
  [1 2]
  [1 2]
  [1 2]]

 [[1 2]
  [1 2]
  [1 2]
  [1 2]]]
(3, 4, 2)