sweep of tensor products

Hello,

I have two matrces A of dimension c x n and B of dimension c x l with c >> n and c >> l

I would like to calculate the tensor C of dimesion c x (n x l) wht C_i,j,k = A_i_j * B_i_k

Is there a way to imrpove the following code ?

Thank you in advance,

c = 100;
n = 5;
l = 10;
A = rand(c, n);
B = rand(c, l);
C = nan([ c n l] );
for ii = 1:c
    C(ii,:,:) = tensorprod(A(ii,:),B(ii,:));
end

Answer 1

[Bruno Luong]

Strictly speaking you do not compute a tensor product.

c = 100;
n = 5;
l = 10;
A = rand(c, n);
B = rand(c, l);

C = A .* reshape(B, c, 1, l);
size(C)

ans = 1×3

100 5 10

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

[Olivier]

Thank you. It is neat and much quicker.

[Matt J]

Strictly speaking you do not compute a tensor product.

It is essentially, a Khatri-Rao product.

[Olivier]

Actually, my original intent is to calculate this (Einstein notation)

C_j,k,l = A_i,j * B_i,k * C_i,l

I calculated first B_i,k * C_i,l and then proceeded to a tensor product.