Sparse matrix from the columns of an initial matrix

Hello everyone, given a matrix A of size (m,n),

I would like to construct a matrix B of size (m,m*n) the following way :

for j = 1:n
    col_start = (j-1)*m+1;
    col_end = j*m;
    B(:,col_start:col_end) = diag(A(:,j));
end

This version uses a for loop, is there any faster way of constructing B?

Thank you in advance.

Answer 1

[Bjorn Gustavsson]

Something like this might work:

vals = []; 
idx1 = [];                      
idx2 = [];
for j = 1:n
idx1 = [idx1,1:n];
  col_start = (j-1)*m+1;                
  col_end = j*m;                        
  idx2 = [idx2,col_start:col_end];
  vals = [vals,A(:,j)'];
end
B2 = sparse(idx1,idx2,vals);

For optimal speed-improvements pre-allocate vals, idx1 and idx2 and assign those with indices instead of growing these arrays.

HTH