Python : How can I find number of 1's in CSR (Compressed Sparse Row matrix) matrix ?

Question

I have created a CSR matrix using scipy.sparse.csr_matrix and the elements in the matrix are either 0 or 1. The matrix is very sparse and big. Is there any way to count how many matrix elements have value 1?

Answer 1

There are several ways to count number of 1's in your CSR matrix, but the quickest way is to use __getattr__

Example:

>>> X.toarray()
array([[0, 1, 0, 0, 1],
       [1, 0, 0, 0, 0],
       [0, 0, 1, 1, 1],
       [1, 0, 0, 1, 0],
       [0, 0, 0, 0, 1],
       [0, 1, 1, 1, 0],
       [1, 0, 0, 0, 0]], dtype=int8)
>>> print(X.__getattr__)
<bound method spmatrix.__getattr__ of <7x5 sparse matrix of type '<class 'numpy.int8'>'
        with 13 stored elements in Compressed Sparse Row format>>
 

Another approach: You can convert the matrix to COOrdinate format using tocoo() function and then check the length.

>>> X_1 = (X == 1)
>>> X_1
<7x5 sparse matrix of type '<class 'numpy.bool_'>'
        with 13 stored elements in Compressed Sparse Row format>
>>> co=X_1.tocoo()
>>> co
<7x5 sparse matrix of type '<class 'numpy.bool_'>'
        with 13 stored elements in COOrdinate format>
>>> m=set(zip(co.row,co.col))
>>> len(m)
13