我有两个具有相同形状的矩阵数组A和B:
A.形状= B.形状=(M,N,P)
我想沿着轴0计算Kronecker乘积,以便:
KP[ii,:,:] = A[ii,:,:]⊗B[ii,:,:]
有没有办法在numpy中执行此操作而不使用for循环?
谢谢!
例:
A = np.array([ [[1,0],
[0,1]],
[[1,0],
[0,1]]
])
B = np.array([ [[1,0],
[0,-1]],
[[0,1],
[1,0]]
])
KP = np.array( [
[[1,0,0,0],
[0,-1,0,0],
[0,0,1,0],
[0,0,0,-1]],
[[0,1,0,0],
[1,0,0,0],
[0,0,0,1],
[0,0,1,0]]
] )
相当于:
KP= np.zeros( (A.shape[0],
A.shape[1]**2,
A.shape[2]**2) )
for ii in range(A.shape[0]):
KP[ii,:,:] = np.kron(A[ii,:,:],B[ii,:,:])
最佳答案
您可以使用einsum
,只需稍加练习即可,它非常直观,也可以采用经典的重塑和广播路线
A = np.array([ [[1,0],
[0,1]],
[[1,0],
[0,1]]
])
B = np.array([ [[1,0],
[0,-1]],
[[0,1],
[1,0]]
])
i,j,k = A.shape
i,l,m = B.shape
np.einsum("ijk,ilm->ijlkm",A,B).reshape(i,j*l,k*m)
# array([[[ 1, 0, 0, 0],
# [ 0, -1, 0, 0],
# [ 0, 0, 1, 0],
# [ 0, 0, 0, -1]],
#
# [[ 0, 1, 0, 0],
# [ 1, 0, 0, 0],
# [ 0, 0, 0, 1],
# [ 0, 0, 1, 0]]])
等效的非einsum表达式:
(A[:,:,None,:,None]*B[:,None,:,None,:]).reshape(i,j*l,k*m)
关于python-3.x - 矩阵阵列的Kronecker积,我们在Stack Overflow上找到一个类似的问题:https://stackoverflow.com/questions/57259557/