Trying to create a multidimensional array

I am trying to implement the formula below, where each is an element of the larger ψ vector of length g.

So far I have this

psi = zeros(1,g);
for i = 1:g
    psi(i) = [zeros(h,(i-1)) eye(h) zeros(h, (N-i-h+1))]';
end

but I receive the error

"Unable to perform assignment because the left and right sides have a different number of elements."

which makes sense. Is there a way to do what I am attempting, or am I SOL?

Answer 1

[Dave B]

I'm not 100% sure I follow the goal, the code you've pasted appears to make g matrices, each with h columns and the number of rows is (i-1) + h + max((N-i-h+1),0)...(I think?)

What shape would you like the result to take? To store the matrices separately, you can use a cell array:

h=4;
g=3;
N=10;
psi = cell(1,g);
for i = 1:g
    psi{i} = [zeros(h,(i-1)) eye(h) zeros(h, (N-i-h+1))]';
end
psi

psi = 1×3 cell array
    {10×4 double}    {10×4 double}    {10×4 double}

If you're using h,g,N that produce consistent numbers of rows (as in the above case) you could store this in a 3-d matrix

h=4;
g=3;
N=10;
psi = zeros(N,h,g);
for i = 1:g
    psi(:,:,i) = [zeros(h,(i-1)) eye(h) zeros(h, (N-i-h+1))]';
end
size(psi)

ans = 1×3

10 4 3

[Josh Rizzolo]

Thanks for getting back to me so fast.

Being completely honest: I'm not 100% on what this array means or how it's supposed to look. The code was basically trying to assemble g matrices of the form given. But, the paper I'm pulling the formula from is certainly trying to concatenate them all into a matrix to do math with.

Constants g, N, and h will all fluctuate depending on the context of the function, hence why they aren't predefined in the code (g = N - h + 1 as a quick aside).

I completely spaced on the proper syntax for putting 2D matrices into 3D ones, so I will try making the adjustment in your second suggestion first. I still anticipate problems stemming from the varying dimensions however.