Scilab Website | Contribute with GitLab | Mailing list archives | ATOMS toolboxes
Scilab Online Help
6.0.0 - Русский

Change language to:
English - Français - 日本語 - Português -

Please note that the recommended version of Scilab is 2024.1.0. This page might be outdated.
See the recommended documentation of this function

Справка Scilab >> Sparse Matrix > Sparse Matrix Manipulation > sprand


sparse random matrix


sp=sprand(nrows,ncols,density [,typ])



integer (number of rows)


integer (number of columns)


filling coefficient (density)


character string, "uniform" (default) or "normal"


sparse matrix


sp=sprand(nrows,ncols,density) returns a sparse matrix sp with nrows rows, ncols columns and approximately density*nrows*ncols non-zero entries.

The density parameter is expected to be in the [0,1] interval. If not, it is automatically projected into this interval. Therefore, using a density which is lower than 0 or greater than 1 will generate neither an error, nor a warning: the formula density=max(min(density,1),0) is used.

If typ="uniform" uniformly distributed values on [0,1] are generated. If typ="normal" normally distributed values are generated (mean=0 and standard deviation=1).

The entries of the output matrix are computed from the given distribution function typ. The indices of the non-zeros entries are computed randomly, so that the average number of nonzeros is equal to density. The actual indices values are computed from the exponential distribution function, where the parameter of the distribution function is computed accordingly.


In the following example, we generate a 100x1000 sparse matrix with approximate density 0.001, i.e. with approximately 100*1000*0.001=100 nonzero entries.

// The entries of the matrix are uniform.
// The entries of the matrix are normal.

In the following script, we check that the entries of the matrix have the expected distribution. We use the spget function in order to get the nonzero entries. Then we compute the min, mean and max of the entries and compare them with the limit values.

typ = "normal";
// typ = "uniform";
nrows = 1000;
ncols = 2000;
density = 1/100;
[%inf -%inf 0 %inf 1] // Limit values for "normal"
[nnzs min(v) mean(v) max(v) variance(v)]
[%inf 0 0.5 1 1/12] // Limit values for "uniform"

In the following script, we check that the entry indices, which are also chosen at random, have the correct distribution. We generate kmax sparse random matrices with uniform distribution. For each matrix, we consider the indices of the nonzero entries which were generated, i.e. we see if the event Aij = {the entry (i,j) is nonzero} occurred for each i and j, for i=1,2,...,nrows and j=1,2,...,ncols. The matrix C(i,j) stores the number of times that the event Aij occurred. The matrix R(k) stores the actual density of the try number k, where k=1,2,...,kmax.

kmax = 1000;
for k=1:kmax
  NZratio = size(NZ,"*")/(nrows*ncols);
  R=[R NZratio];

Now that this algorithm has been performed (which may require some time), we can compute elementary statistics to check that the algorithm performed well.

// The expectation of A_ij is
density * kmax
// Compare this with the actual events :
// The average number should be close to the expectation.
[density*kmax mean(C)]
// The density should be close to expected density
[density mean(R)]

See also

  • sparse — sparse matrix definition
  • full — sparse to full matrix conversion
  • rand — Случайные числа
  • speye — sparse identity matrix
Report an issue
<< spones Sparse Matrix Manipulation spzeros >>

Copyright (c) 2022-2024 (Dassault Systèmes)
Copyright (c) 2017-2022 (ESI Group)
Copyright (c) 2011-2017 (Scilab Enterprises)
Copyright (c) 1989-2012 (INRIA)
Copyright (c) 1989-2007 (ENPC)
with contributors
Last updated:
Tue Feb 14 15:13:25 CET 2017