How do you extract diagonal elements of submatrix matrix?

Hi, need help:

I have a (nxn) square matrix A. I then created submatrices of (kxk) within matrix A. How do I extract only the diagonal (kxk) submatrices into a separate vector? The command diag(A) does not work; this is maybe because of the submatrices elements that I created. Please advise Thanks!

Answer 1

[Ameer Hamza]

Are the submatrices in cell array? For example, try something like this

n = 9;
A = rand(n);

k = 3;
B = mat2cell(A, k*ones(1,n/k), k*ones(1,n/k));

diag_all = cellfun(@diag, B, 'uni', 0);

[BooDee]

This works well! Thanks Ameer!

[Ameer Hamza]

I am glad to be of help!

[BooDee]

Ameer,

I am also interested in the diagonal cells (or blocks) of the B matrix. I tried using the following code as such:

diag_B = cellfun(@diag, B);

But it didn't work.

After reading about 'cellfun', this might not be the right funciton for me to extract the information I need because the output of 'cellfun' is an array. And the diagonal of B would be submatrices. I hope you understand my point.

Please advise. Thanks so much in advance!

[Ameer Hamza]

Can you further explain? Is B a cell array? Have you tried with 'uni', 0 option as shown in my answer?

[BooDee]

Hi Ameer,

Sorry for the late response.

No, B is not a cell array. It is the same B you simulated, which is in it contains blocks of 3x3 matrices.

I wanted to get the diagonal blocks of that B. I've tried with uni, 0 option, didn't work either.

I also tried using the command:

diag_B = blkdiag(B{:}) but that didn't work either.

Please advise. Thanks a lot Ameer!

[Ameer Hamza]

But this line already extract the diagonal elements

diag_all = cellfun(@diag, B, 'uni', 0);

Can you post the error message?

[BooDee]

Hi Ameer,

Yes, you are right that functon:

diag_all = cellfun(@diag, B, 'uni', 0);

gives you the diagonal elements of the diagonal blocks of B. So, that's done. I am so grateful for your solution.

My follow up question is how can i get JUST the diagonal blocks of B (in your simulation, thats the 3x3 matrices) and separate them as a separate and individual matrices. The logical function would be to use blkdiag(B) but that didn't seem to work.

Again, thanks Ameer!

[Ameer Hamza]

Try this

diag_all = cellfun(@diag, B, 'uni', 0);
diag_B = blkdiag(diag_all{:})

[BooDee]

Hey Ameer,

Thanks for the response. I've done the codes you suggested above but to no avail. It keeps giving me the following errors:

Error using zeros

Out of memory. Type HELP MEMORY for your options.

Error in blkdiag (line 43)

y = zeros(p1(end),m1(end),outclass); %Preallocate

I tried just simplifying it with:

diag_B = blkdiag(B{:}) AND diag_B = diag(B);

But it still won't work.

Is there any other solution. I can't imagine doing this manually when working with a 2464x2464 matrix.

Thanks Ameer. Anything helps.

[Vishnu Pradeep]

Hey guys,

What does the 'uni' argument pertain to?

Couldnt find anything on the web other than the name-value pair 'UniformOutput'. Is this the same?

Thanks!