NumPy and SciPy – Difference between .todense() and .toarray()
Question:
I am wondering if there is any difference (advantage/disadvantage) of using .toarray() vs. .todense() on sparse NumPy arrays. E.g.,
import scipy as sp
import numpy as np
sparse_m = sp.sparse.bsr_matrix(np.array([[1,0,0,0,1], [1,0,0,0,1]]))
%timeit sparse_m.toarray()
1000 loops, best of 3: 299 µs per loop
%timeit sparse_m.todense()
1000 loops, best of 3: 305 µs per loop
Answers:
toarray returns an ndarray; todense returns a matrix. If you want a matrix, use todense; otherwise, use toarray.
As its documentation states, using numpy.matrix is discouraged as it may be removed in the future. So it’s probably better to use toarray() rather than todense(), especially since the matrix operations that were convenient on matrix objects are now possible on ndarray objects since numpy 1.10.
I am wondering if there is any difference (advantage/disadvantage) of using .toarray() vs. .todense() on sparse NumPy arrays. E.g.,
import scipy as sp
import numpy as np
sparse_m = sp.sparse.bsr_matrix(np.array([[1,0,0,0,1], [1,0,0,0,1]]))
%timeit sparse_m.toarray()
1000 loops, best of 3: 299 µs per loop
%timeit sparse_m.todense()
1000 loops, best of 3: 305 µs per loop
toarray returns an ndarray; todense returns a matrix. If you want a matrix, use todense; otherwise, use toarray.
As its documentation states, using numpy.matrix is discouraged as it may be removed in the future. So it’s probably better to use toarray() rather than todense(), especially since the matrix operations that were convenient on matrix objects are now possible on ndarray objects since numpy 1.10.