CODEWITHSUNDEEP
pythonnumpyscipysparse-matrix
Kam
Kam

Reputation: 87

AttributeError using scipy.sparse.toarray()

import scipy as sp
import numpy as np

a=sp.sparse.coo_matrix(np.random.randint(0,9,[4,5]))
b=sp.sparse.coo_matrix(np.random.randint(0,9,[4,2]))
sp.hstack([a,b]).toarray()

is giving me

AttributeError: 'numpy.ndarray' object has no attribute 'toarray'

could you help me with my silly mistake here?

Upvotes: 1

Views: 4764

Answers (1)

DSM
DSM

Reputation: 353179

sp.hstack (i.e. numpy.hstack) is the ordinary, dense hstack, which won't combine the sparse arrays correctly. It builds a 1D numpy array already (of object dtype; in other words, it just wraps the Python-level objects and crams them in there.) You want scipy.sparse.hstack:

In [332]: sp.hstack([a, b])
Out[332]: 
array([<4x5 sparse matrix of type '<class 'numpy.int64'>'
    with 17 stored elements in COOrdinate format>,
       <4x2 sparse matrix of type '<class 'numpy.int64'>'
    with 7 stored elements in COOrdinate format>], dtype=object)

In [333]: sp.hstack([a, b]).toarray()
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-333-4f469eb98b12> in <module>()
----> 1 sp.hstack([a, b]).toarray()

AttributeError: 'numpy.ndarray' object has no attribute 'toarray'

In [334]: sp.sparse.hstack([a, b])
Out[334]: 
<4x7 sparse matrix of type '<class 'numpy.int64'>'
    with 24 stored elements in COOrdinate format>

In [335]: sp.sparse.hstack([a, b]).toarray()
Out[335]: 
array([[3, 2, 7, 0, 5, 5, 1],
       [7, 1, 2, 1, 7, 0, 8],
       [6, 1, 6, 1, 8, 6, 2],
       [7, 6, 0, 5, 0, 8, 8]], dtype=int64)

Upvotes: 2

Related Questions