如何将整数值稀疏矩阵变换为0/1值稀疏矩阵，Python

Question

I have a sparse matrix from the sklearn bag-of-words vectorizer. It's a csr_matrix and its elements represent word frequency in a document. But now what I need is the 0/1 matrix where 1 represents the word exists in the document, so I don't care about the actual frequency. Disregard the background problem, it's like this: I have a sparse matrix,

2 3 4 0 0 0
0 0 0 0 0 8
0 0 0 2 0 0
0 0 0 0 0 0

I want all the nonzero elements to be 1,

1 1 1 0 0 0
0 0 0 0 0 1
0 0 0 1 0 0
0 0 0 0 0 0

How can I achieve this? I assume using todense() and then loop is not a good choice since the sparse matrix is large. Is there a better way?

Answer 1

Try csr_matrix.sign . it should be exactly what you need (although I didn't try it myself).

Answer 2

I think you could just create a new matrix from the non-zero indices (see the scipy.sparse.csr_matrix reference). Assuming your sparse matrix is named sp_m:

sp_unit = csr_matrix( ([1]*len(sp_m.data), sp_m.nonzero()), shape=sp_m.shape )

OR

As another user pointed out, you could use the sign function; however, I think you will need to square it if you have negative values:

sp_unit = sp_m.sign.multiply(sp_m.sign)