Converting a 2d array into a 3d array

Hi,

I am having a 2d array with size mxn

I want to convert each column into a seperate 2d diagonal matrix--- a total of 'n' diagonal matrices of size mxm

Then combining all the 2d diagonal matrices as a single 3d matrix.---size mxmxn

I tried to do this with employing a for-loop but it is taking lot of time

Can this be done without using a loop and at a faster speed?

[Turlough Hughes]

Can you give an example of the code you have implemented?

[Vinay Killamsetty]

size of Z_int is mxn

Z = nan(m,m,n)

for i = 1 : size( Z_int , 2)

Z( : , : , i ) = diag( (Z_int( :, i ) ) ;

end

Answer 1

[Chunru]

"it is taking lot of time" for what size of array?

m=200; n=100;
Z_int = rand(m, n);
Z = nan(m,m,n);
tic
for i = 1 : size( Z_int , 2)
    Z( : , : , i ) = diag( Z_int( :, i ) ); 
end
toc

Elapsed time is 0.009776 seconds.

Answer 2

[Turlough Hughes]

You could write a linear index for the diagonal elements as follows:

[m,n] = size(Z_int)
Z = zeros(m,m,n);

idx = repmat((1:m+1:m^2).',1,n) + m^2*(0:n-1);
Z(idx(:)) = Z_int(:)

This is 2-4x faster for m >= 400 on my computer (n=500), however, on the matlab servers its a slightly slower. Whether it's faster or not for you will depend on your hardware, the size of your arrays and probably the version of matlab that you're using too.