Array multiplication along certain dimension?

60 views (last 30 days)
I have a 4-D array with dimensions A = A1 x A2 x A3 x N and a vector with dimensions V = 1 x N. I would like to multiply each element in the array A with vector V along fourth dimension and take a sum along the same dimension to form a matrix B = A1 x A2 x A3. Currently I am using a loop to achieve this, but is there any faster method to achieve the same result? Thanks!
A1 = 3;
A2 = 3;
A3 = 3;
N = 10;
A = ones(A1, A2, A3, N);
V = ones(1, N);
B = NaN(A1, A2, A3);
for i = 1:A1
for j = 1:A2
for k = 1:A3
B(i,j,k) = squeeze(A(i,j,k,:))' * V';
end
end
end

Accepted Answer

David Goodmanson
David Goodmanson on 24 Nov 2017
Edited: David Goodmanson on 24 Nov 2017
Hi Visa,
This seems to work
A1 = 2;
A2 = 3;
A3 = 4;
N = 7;
A = rand(A1, A2, A3, N);
V = rand(1, N);
% ---- method 1
B1 = NaN(A1, A2, A3);
for i = 1:A1
for j = 1:A2
for k = 1:A3
B1(i,j,k) = squeeze(A(i,j,k,:))' * V';
end
end
end
% ---- method 2
D = permute(A,[4 1 2 3]);
E = reshape(D,N,A1*A2*A3);
F = V*E;
B2 = reshape(F,A1,A2,A3);
max(max(max(max(abs(B1-B2))))) ans = 8.8818e-16

More Answers (1)

James Tursa
James Tursa on 27 Nov 2017
Another way:
B = sum(bsxfun(@times,A,reshape(V,1,1,1,numel(V))),4);
Or in later versions of MATLAB:
B = sum(A.*reshape(V,1,1,1,numel(V)),4);
  1 Comment
Visa Suomi
Visa Suomi on 28 Nov 2017
Thanks, they work with smaller grids, but I encounter memory problems with my array sizes with these functions. My array A = 120 x 132 x 132 x 120.

Sign in to comment.

Categories

Find more on Matrices and Arrays in Help Center and File Exchange

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!