[FIXED] Matrix multiplication for numpy array of different dimensions

Issue

Assume we have 6 vectors (each vector has 1*4 dimension) in a numpy array having shape (3, 2, 4) like this:

A=np.array([   [[1,2,3,4], [5,6,7,8]],    [[ 7,8 ,9,10],[ 11,12 ,13,14]],    [[ 15,16 ,17,18],[ 19,20 ,21,22]]   ])

Assume we have single Matrix B having dimensions 4*4 in a np.array form having (4, 4) shape as follows

B=np.array([  [1,2,3,4],     [5,6,7,8],      [9,10,11,12], [13,14,15,16]])

How to perform matrix multiplication of matrix B to every vector of the 6 vectors in A and the result have the same shape as A the resalut like this

result=np.array([   [[1,2,3,4]@B, [5,6,7,8]@B],    [[ 7,8 ,9,10]@B,[ 11,12 ,13,14]@B],    [[ 15,16 ,17,18]@B,[ 19,20 ,21,22]@B]   ])

result=np.array([[[ 90, 100, 110, 120],
        [202, 228, 254, 280]],

       [[258, 292, 326, 360],
        [370, 420, 470, 520]],

       [[482, 548, 614, 680],
        [594, 676, 758, 840]]])

Solution

Edit:

It seems you are looking for np.matmul:

>>> np.matmul(A, B)
array([[[ 90, 100, 110, 120],
        [202, 228, 254, 280]],

       [[258, 292, 326, 360],
        [370, 420, 470, 520]],

       [[482, 548, 614, 680],
        [594, 676, 758, 840]]])

---

However, you can probably use numpy broadcasting to get what you expect:

>>> Stu_data[:, :, None] * A_data

array([[[[   1,    4,    9,   16],
         [   5,   12,   21,   32],
         [   9,   20,   33,   48],
         [  13,   28,   45,   64]],

        [[  55,  132,  231,  352],
         [ 275,  396,  539,  704],
         [ 495,  660,  847, 1056],
         [ 715,  924, 1155, 1408]]],


       [[[   1,    4,    9,   16],
         [   5,   12,   21,   32],
         [   9,   20,   33,   48],
         [  13,   28,   45,   64]],

        [[   1,    4,    9,   16],
         [   5,   12,   21,   32],
         [   9,   20,   33,   48],
         [  13,   28,   45,   64]]],


       [[[   1,    4,    9,   16],
         [   5,   12,   21,   32],
         [   9,   20,   33,   48],
         [  13,   28,   45,   64]],

        [[   1,    4,    9,   16],
         [   5,   12,   21,   32],
         [   9,   20,   33,   48],
         [  13,   28,   45,   64]]]])

Shapes:

>>> A_data
(4, 4)

>>> Stu_data
(3, 2, 4)

>>> Stu_data[:, :, None].shape
(3, 2, 1, 4)

>>> (Stu_data[:, :, None] * A_data).shape
(3, 2, 4, 4)

Answered By - Corralien