5.6.2 Generating sparse matrices

The NumPy commands eye, identity, diag, and rand have their sparse counterparts. They take an additional argument; it specifies the sparse matrix format of the resulting matrix.

The following commands generate the identity matrix but in different sparse matrix formats:

import scipy.sparse as sp
sp.eye(20,20,format = 'lil') 
sp.spdiags(ones((20,)),0,20,20, format = 'csr') 
sp.identity(20,format ='csc')

The command sp.rand takes an additional argument describing the density of the generated random matrix. A dense matrix has density 1 while a zero matrix has density 0:

import scipy.sparse as sp 
AS=sp.rand(20,200,density=0.1,format='csr')
AS.nnz # returns 400

There is no direct correspondence to the NumPy command zeroes. Matrices completely filled with zeros are generated by instantiating the corresponding type with the shape parameters as constructor parameters...