Extract rows in 3D matrix according to indexes

x2 = zeros(60, 46000, 'single');
for N = 1:46000
x2(:,N) = x(:,i(N),N);
end

Hi. I have a 3D matrix x with the dimensions 60x100x40000 and a 1x46000 vector i with an index for each matrix in the 3rd dimension telling me which columns I need to extract (The values for i are between 1 & 100). So ultimately, I would like to end up with a 60x40000 matrix, using i to extract from x.

My solution so far can be seen above. It works, but takes forever (especially since I have to repeat the process almost 100 times. Does anyone know of a direct way I can do this?

Thanks!

Answer 1

[Andrei Bobrov]

% Let X - your data with size  (60x100x40000)
% ii - your vector with index of columns X from each frame of X (1x46000)

[m,n,k] = size(X);
x1 = reshape(X,m,[]);
x2 = x1(:,n*(0:k-1)+ii(1:k));

Answer 2

[Stephen23]

Perhaps you could try the values of the index vector directly:

% index:
idx = [2,3,1,2];
% fake data:
X = reshape(1:2*3*4,2,3,4);
% extract columns:
Y = X(:,idx,:);
% create index for pages:
[R,C,~] = size(Y);
idy = cumsum([1:R;R*(C+1)*ones(C-1,2)]);
% extract pages:
A = Y(idy).';

This gives the identical results to your loop, but with the advantange that it does not use any loop.