CODEWITHSUNDEEP
pythonnumpymatrixscipysparse-matrix
Daniel Mahler
Daniel Mahler

Reputation: 8203

efficiently set a large number of SciPy sparse matrix entries to zero

I need to prune a large number of entries from a SciPy sparse matrix. Currently I convert the matrix to the DOK format and individually assign each entry to 0.

m = m.todok()
for i, j in pruneme:
  m[i,j] = 0

This is extremely slow.

Is there a faster way?

Upvotes: 3

Views: 207

Answers (1)

Roland W
Roland W

Reputation: 1461

You can set elements of CSR sparse arrays efficiently, as long as you do not add new nonzeros, simply by subscripting the array with tuples:

i, j = zip(*pruneme) # assuming that pruneme is a python list
m[i, j] = 0.
m.eliminate_zeros()

That should be much faster than constructing two arrays.

Upvotes: 4

Related Questions