Construct a block sparse matrix. The meaning of arguments is analogous to the
built-in sparse function, except that i, j are indices of
blocks rather than elements, and sv is a 3-dimensional array, the first two
dimensions determining the block size. Optionally, m and n can be
specified as the true block dimensions; if not, the maximum values of i, j
are taken instead. The resulting sparse matrix has the size
[m*p, n*q]
where
p = size (sv, 1) q = size (sv, 2)
The blocks are located so that
s(i(k):i(k)+p-1, j(k):j(K)+q-1) = sv(:,:,k)
Multiple blocks corresponding to the same pair of indices are summed, unless mode is "unique", in which case the last of them is used.
Package: linear-algebra