Find all rows in a CSR matrix with only 0 in all columns

Question

I want to find all rows in a Compressed Sparse Row (CSR) matrix that contain only zero values. Is there any scipy function that I can use to find?

Answer 1

You can use the count_nonzero() method to count the number of non-zero entries in each row of a CSR matrix. Rows that return 0 will have all columns 0.

Here is an example:

---

import numpy as np

from scipy.sparse import csr_matrix

# data
row = np.array([0, 0, 2, 2, 2, 3])
col = np.array([0, 2, 0, 1, 2, 2])
data = np.array([1, 2, 3, 4, 5, 6])

# Create CSR matrix
csr_data = csr_matrix((data, (row, col)), shape=(4, 4))

zero_rows = []
for i in range(csr_data.shape[0]):
    if csr_data[i].count_nonzero() == 0:
        zero_rows.append(i)

print("rows with only 0: ", zero_rows)

---

The above code will print [1] as row 1 has all columns 0.