MATLAB®never createssparse matrices automatically. Instead, you must determine if a matrix contains a large enough percentage of zeros to benefit from sparse techniques.
Thedensityof a matrix is the number of nonzero elements divided by the total number of matrix elements. For matrixM
, this would be
nnz(M) / prod(size(M));
nnz(M) / numel(M);
Matrices with very low density are often good candidates for use of the sparse format.
You can convert a full matrix to sparse storage using thesparse
function with a single argument.
S = sparse(A)
For example:
A = [ 0 0 0 5 0 2 0 0 1 3 0 0 0 0 4 0]; S = sparse(A)
produces
S = (3,1) 1 (2,2) 2 (3,2) 3 (4,3) 4 (1,4) 5
The printed output lists the nonzero elements ofS
, together with their row and column indices. The elements are sorted by columns, reflecting the internal data structure.
You can convert a sparse matrix to full storage using thefull
function, provided the matrix order is not too large. For example,A = full(S)
reverses the example conversion.
Converting a full matrix to sparse storage is not the most frequent way of generating sparse matrices. If the order of a matrix is small enough that full storage is possible, then conversion to sparse storage rarely offers significant savings.
You can create a sparse matrix from a list of nonzero elements using thesparse
function with five arguments.
S = sparse(i,j,s,m,n)
i
andj
are vectors of row and column indices, respectively, for the nonzero elements of the matrix.s
is a vector of nonzero values whose indices are specified by the corresponding(i,j)
pairs.m
is the row dimension of the resulting matrix, andn
is the column dimension.
The matrixS
of the previous example can be generated directly with
S = sparse([3 2 3 4 1],[1 2 2 3 4],[1 2 3 4 5],4,4) S = (3,1) 1 (2,2) 2 (3,2) 3 (4,3) 4 (1,4) 5
Thesparse
command has a number of alternate forms. The example above uses a form that sets the maximum number of nonzero elements in the matrix tolength(s)
. If desired, you can append a sixth argument that specifies a larger maximum, allowing you to add nonzero elements later without reallocating the sparse matrix.
The matrix representation of the second difference operator is a good example of a sparse matrix. It is a tridiagonal matrix with -2s on the diagonal and 1s on the super- and subdiagonal. There are many ways to generate it—here's one possibility.
D = sparse(1:n,1:n,-2*ones(1,n),n,n); E = sparse(2:n,1:n-1,ones(1,n-1),n,n); S = E+D+E'
Forn = 5
, MATLAB responds with
S = (1,1) -2 (2,1) 1 (1,2) 1 (2,2) -2 (3,2) 1 (2,3) 1 (3,3) -2 (4,3) 1 (3,4) 1 (4,4) -2 (5,4) 1 (4,5) 1 (5,5) -2
NowF = full(S)
displays the corresponding full matrix.
F = full(S) F = -2 1 0 0 0 1 -2 1 0 0 0 1 -2 1 0 0 0 1 -2 1 0 0 0 1 -2
Creating sparse matrices based on their diagonal elements is a common operation, so the functionspdiags
handles this task. Its syntax is
S = spdiags(B,d,m,n)
To create an output matrixS
的大小m-by-nwith elements onp
diagonals:
B
is a matrix of sizemin(m,n)
-by-p. The columns ofB
are the values to populate the diagonals ofS
.
d
is a vector of lengthp
whose integer elements specify which diagonals ofS
to populate.
That is, the elements in columnj
ofB
fill the diagonal specified by elementj
ofd
.
If a column ofB
is longer than the diagonal it's replacing, super-diagonals are taken from the lower part of the column ofB
, and sub-diagonals are taken from the upper part of the column ofB
.
As an example, consider the matrixB
and the vectord
.
11 B = [41 0 52 22 0 63 33 74 44 24);d = [30 2];
Use these matrices to create a 7-by-4 sparse matrixA
:
A = spdiags(B,d,7,4) A = (1,1) 11 (4,1) 41 (2,2) 22 (5,2) 52 (1,3) 13 (3,3) 33 (6,3) 63 (2,4) 24 (4,4) 44 (7,4) 74
In its full form,A
looks like this:
full(A) ans = 11 0 13 0 0 22 0 24 0 0 33 0 41 0 0 44 0 52 0 0 0 0 63 0 0 0 0 74
spdiags
can also extract diagonal elements from a sparse matrix, or replace matrix diagonal elements with new values. Type帮助
spdiags
for details.
You can import sparse matrices from computations outside the MATLAB environment. Use thespconvert
function in conjunction with theload
command to import text files containing lists of indices and nonzero elements. For example, consider a three-column text fileT.dat
whose first column is a list of row indices, second column is a list of column indices, and third column is a list of nonzero values. These statements loadT.dat
into MATLAB and convert it into a sparse matrixS
:
load T.dat S = spconvert(T)
Thesave
andload
commands can also process sparse matrices stored as binary data in MAT-files.