Hi all,
Since I'm working with very large and sparse matrices I use sparse() to define these matrices. I do this by defining the index of the rows and columns and their corresponding value in separate vectors. For example, for a row i and columns j,j-1,j+1, the values in the matrix are defined according:
s = 1;row(s) = i; column(s)=j; K(s) = 1; s=s+1;row(s) = i; column(s)=j-1; K(s) = -1; s=s+1;row(s) = i; column(s)=j+1; K(s) = -1; s=s+1;%This is done for all non-zero points
K_matrix = sparse(row,column,K);
This works great when specifying distinct values. However, how can I use this sparse indexing when the value depends on values in other rows of the matrix. For example:
K(i,:) = K(i-1,:) + K(i+1,:);
i.e. I need all the values of another row in order to construct the current row, since I can't do something like this:
row(s) = i; column(s)=:; K(s) = K(i-1,:)+K(i+1,:); s=s+1;
I use this way of defining the sparse matrix since it is recommended for speed. Does anyone know how I can circumvent this problem without Matlab becoming slow when building the matrix?
Kind regards
Best Answer