我有矩阵 (shape=3,12)并想用向量进行列到列的乘法 kkk (len=6) .我可以通过制作 np.tile(kkk,2) 来做到这一点将其扩展为 12 元素向量。但我想知道有没有一种方法可以直接将它相乘而不使用 np.tile (或者我必须为乘法制作一个新向量的任何东西)?
>>> matrix=np.ones([3,12])
>>> kkk
array([ 0.008, 0.595, 0.278, 0.103, 0.014, 0.002])
>>> matrix*np.tile(kkk,2)
array([[ 0.008, 0.595, 0.278, 0.103, 0.014, 0.002, 0.008, 0.595, 0.278, 0.103, 0.014, 0.002],
[ 0.008, 0.595, 0.278, 0.103, 0.014, 0.002, 0.008, 0.595, 0.278, 0.103, 0.014, 0.002],
[ 0.008, 0.595, 0.278, 0.103, 0.014, 0.002, 0.008, 0.595, 0.278, 0.103, 0.014, 0.002]])
请您参考如下方法:
reshape 以将 matrix 的第二个轴分成两部分,最后一个轴的长度为 6,然后我们可以将其引入 kkk 的逐元素乘法 最后 reshape 回原来的形状 -
(matrix.reshape(3,2,6)*kkk).reshape(3,12)
sample 运行-
In [47]: a = np.random.randint(0,9,(3,12))
In [48]: matrix = np.random.randint(0,9,(3,12))
In [49]: kkk = np.random.randint(0,9,(6))
# As proposed in this post
In [50]: (matrix.reshape(3,2,6)*kkk).reshape(3,12)
Out[50]:
array([[21, 32, 1, 5, 8, 6, 12, 12, 2, 0, 16, 9],
[18, 12, 7, 8, 40, 24, 3, 12, 1, 2, 48, 15],
[15, 20, 1, 0, 16, 12, 3, 32, 7, 8, 48, 24]])
# As proposed in the question with tiling (for verification)
In [52]: matrix*np.tile(kkk,2)
Out[52]:
array([[21, 32, 1, 5, 8, 6, 12, 12, 2, 0, 16, 9],
[18, 12, 7, 8, 40, 24, 3, 12, 1, 2, 48, 15],
[15, 20, 1, 0, 16, 12, 3, 32, 7, 8, 48, 24]])
