Which is more efficient when applied to a sparse matrix, find or logical indexing ?

Given a sparse matrix, I would like to get its non-zero entries. I would like to know if using find function is an efficient way to do this or are there more efficient ways. Moreover, once I have a sparse matrix S in [rows, cols, vals] format with size nnz(S), how much cost is associated with executing sparse(rows, cols, vals, m, n). Thank you ...

[KSSV]

logical indexing....you can check yourself by using tic toc.

Answer 1

[Bruno Luong]

For sparse matrix using FIND to get non-zero elements is the best - bar none. It's design for it and the data are internally ranged such a way that MATLAB returns index/value effortlessly. Just do it directly on the matrix:

[rows, cols, vals] = find(S);

(However if you to find with some addition logical operator (unitary/binary) on the matrix, this is another matter.)

The cost of

sparse(rows, cols, vals, m, n)

is quite reasonable, and it's indeed the best way to build sparse matrix. Avoid to update elements of a prebuilt sparse matrix iteratively. The later method is quite innefficient.