accessing elements through sparse matrix in scipy - python-3.x

Question

I have below code in python
# dense to sparse
from numpy import array
from scipy.sparse import csr_matrix
# create dense matrix
A = array([[1, 0, 0, 1, 0, 0], [0, 0, 2, 0, 0, 1], [0, 0, 0, 2, 0, 0]])
print(A)
# convert to sparse matrix (CSR method)
S = csr_matrix(A)
print(S)
# reconstruct dense matrix
B = S.todense()
print(B)
Above code when I have following statement I have
print(B[0])
I have following output:
[[1 0 0 1 0 0]]
How can I loop through the above values i.e, 1, 0, 0, 1, 0, 0, 0

Answer 1

In [2]: from scipy.sparse import csr_matrix
...: # create dense matrix
...: A = np.array([[1, 0, 0, 1, 0, 0], [0, 0, 2, 0, 0, 1], [0, 0, 0, 2, 0, 0]])
...: S = csr_matrix(A)
In [3]: A
Out[3]:
array([[1, 0, 0, 1, 0, 0],
[0, 0, 2, 0, 0, 1],
[0, 0, 0, 2, 0, 0]])
In [4]: S
Out[4]:
<3x6 sparse matrix of type '<class 'numpy.int64'>'
with 5 stored elements in Compressed Sparse Row format>
S.toarray() or S.A for short, makes a dense ndarray:
In [5]: S.A
Out[5]:
array([[1, 0, 0, 1, 0, 0],
[0, 0, 2, 0, 0, 1],
[0, 0, 0, 2, 0, 0]])
todense makes a np.matrix object, which is always 2d
In [6]: S.todense()
Out[6]:
matrix([[1, 0, 0, 1, 0, 0],
[0, 0, 2, 0, 0, 1],
[0, 0, 0, 2, 0, 0]])
In [7]: S.todense()[0]
Out[7]: matrix([[1, 0, 0, 1, 0, 0]])
In [9]: S.todense()[0][0]
Out[9]: matrix([[1, 0, 0, 1, 0, 0]])
To iterate by 'columns' we have to do something like:
In [10]: [S.todense()[0][:,i] for i in range(3)]
Out[10]: [matrix([[1]]), matrix([[0]]), matrix([[0]])]
In [11]: [S.todense()[0][0,i] for i in range(3)]
Out[11]: [1, 0, 0]
There is a shortcut for converting a 1d row np.matrix to a 1d ndarray:
In [12]: S.todense()[0].A1
Out[12]: array([1, 0, 0, 1, 0, 0])
Get a 1d array from a "row" of a ndarray is simpler:
In [14]: S.toarray()[0]
Out[14]: array([1, 0, 0, 1, 0, 0])
np.matrix is generally discouraged, as a remnant from a time when the transition from MATLAB was more important. Now that fact that sparse is modeled on np.matrix (but not subclassed) is the main reason for keeping np.matrix. Row and column sums of a sparse matrix return dense matrix.