3d Matrix - Extract Subarray and Multiply by Conjugate Transpose without Forloops

3 次查看（过去 30 天）
Science Machine
Science Machine 2022-9-6
编辑： Science Machine 2022-9-8
I have a 3d matrix A(i,j,k) of size [1:100,1:10000,1:989]. On the kth index I want to extract the 989 elements into a vector u and form the product u*ctranspose(u), for each of the indices.
Using a double for loop, which is not something one should do in Matlab,
% this is an evil double for loop - I want to avoid doing this.
for ii=1:100
for jj=1:10000
u=squeeze(A(ii,jj,:));
% somehow compute and store uu in a vectorized way, but how?
uu=u*ctranspose(u); % note that uu is a 989x989 matrix, not a vector.
end %ii
end %jj
This would be really slow. Is there a vectorized way to the above, so I am not doing a double for loop?
  17 个评论
显示 15更早的评论
Science Machine
Science Machine 2022-9-8
编辑：Science Machine 2022-9-8
t and ras parameters are the ii and jj indices from above u of size 1:100 and 1:10000, and the vector $$ has parameters jj also.

请先登录，再进行评论。

请先登录，再回答此问题。

采纳的回答

Matt J
Matt J 2022-9-7
编辑：Matt J 2022-9-7
[m,n,p]=size(A);
Ar=reshape(A,[],p).';
UU = reshape(Ar,p,1,[]).*conj(reshape(Ar,1,p,[]));
UU=reshape(UU,p,p,m,n); %obtain the (i,j)-th outer product as UU(:,:,i,j)
  3 个评论
显示 1更早的评论
Science Machine
Science Machine 2022-9-7
How did you learn to do such things? I knew reshape function but it did not occur to me to use it here.

请先登录，再进行评论。

更多回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Correlation and Convolution 的更多信息

标签

产品


版本

R2022a

Community Treasure Hunt

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

Start Hunting!

Translated by