7

Given an arbitrary numpy array (ndarray), is there a function or a short way to convert it to a scipy.sparse matrix?

I'd like something that works like:

A = numpy.array([0,1,0],[0,0,0],[1,0,0])
S = to_sparse(A, type="csr_matrix")
clstaudt
  • 21,436
  • 45
  • 156
  • 239

3 Answers3

9

I usually do something like

>>> import numpy, scipy.sparse
>>> A = numpy.array([[0,1,0],[0,0,0],[1,0,0]])
>>> Asp = scipy.sparse.csr_matrix(A)
>>> Asp
<3x3 sparse matrix of type '<type 'numpy.int64'>'
    with 2 stored elements in Compressed Sparse Row format>
DSM
  • 342,061
  • 65
  • 592
  • 494
  • Hm, that was easy, I should have tried that. I didn't expect the constructor to do the conversion. This works for the constructors of the other sparse matrix types, too. – clstaudt May 21 '12 at 14:28
4

A very useful and pertinent example is in the help!

import scipy.sparse as sp
help(sp)

This gives:

Example 2
---------

Construct a matrix in COO format:

>>> from scipy import sparse
>>> from numpy import array
>>> I = array([0,3,1,0])
>>> J = array([0,3,1,2])
>>> V = array([4,5,7,9])
>>> A = sparse.coo_matrix((V,(I,J)),shape=(4,4))

It's also worth noting the various constructors are (again from the help):

    1. csc_matrix: Compressed Sparse Column format
    2. csr_matrix: Compressed Sparse Row format
    3. bsr_matrix: Block Sparse Row format
    4. lil_matrix: List of Lists format
    5. dok_matrix: Dictionary of Keys format
    6. coo_matrix: COOrdinate format (aka IJV, triplet format)
    7. dia_matrix: DIAgonal format

To construct a matrix efficiently, use either lil_matrix (recommended) or
dok_matrix. The lil_matrix class supports basic slicing and fancy
indexing with a similar syntax to NumPy arrays.

Your example would be as simple as:

S = sp.csr_matrix(A)
Hooked
  • 84,485
  • 43
  • 192
  • 261
  • 1
    @clstaudt I'm not sure what you are looking for, `sp.csr_matrix(A)` would construct a `csr` type matrix from a dense numpy array `A`, while `sp.csc_matrix(A)` would construct a type `csc` matrix, etc... The constructors _are_ the efficient conversion functions you are looking for (though read the caveats in the help). – Hooked May 21 '12 at 14:28
  • You're right, the constructors do the conversions I was looking for. I expected them to take only arguments that correspond to the way entries are stored, like: A = sparse.coo_matrix((V,(I,J)),shape=(4,4)) – clstaudt May 21 '12 at 14:32
0

Please refer to this answer: https://stackoverflow.com/a/65017153/9979257

In this answer, I have explained how to convert a 2-dimensional NumPy matrix into CSR or CSC format.

Yogesh Awdhut Gadade
  • 2,498
  • 24
  • 19