Create CSR matrix from x_index, y_index, value

问题

I have data of the format (x_index, y_index, value) and I'm trying to create a CSR matrix using scipy (scipy.sparse.csr.csr_matrix).

For example, convert:

To the following:

10 5
3  4

I've read the documentation here: http://docs.scipy.org/doc/scipy/reference/generated/scipy.sparse.csr_matrix.html

However I'm still not clear which of the examples applies to my use case.

回答1:

If you can separate the input data into a sequence of row indices, a sequence of column indices and a corresponding sequence of value indices, you can use the fourth option shown in the csr_matrix docstring for creating the matrix.

For example, supposed you already have your data in a single array, data, where the first two columns are the indices and the third column holds the values. E.g.

In [213]: data
Out[213]: 
array([[ 0,  0, 10],
       [ 0,  1,  5],
       [ 1,  0,  3],
       [ 1,  1,  4]])

Then you can create a CSR matrix as follows:

In [214]: a = csr_matrix((data[:, 2], (data[:, 0], data[:, 1])))

In [215]: a
Out[215]: 
<2x2 sparse matrix of type '<type 'numpy.int64'>'
    with 4 stored elements in Compressed Sparse Row format>

In [216]: a.A
Out[216]: 
array([[10,  5],
       [ 3,  4]])

Depending on your data, you might need to specify the shape explicitly. For example, here I use the same data, but in a 3x3 matrix:

In [217]: b = csr_matrix((data[:, 2], (data[:, 0], data[:, 1])), shape=(3, 3))

In [218]: b
Out[218]: 
<3x3 sparse matrix of type '<type 'numpy.int64'>'
    with 4 stored elements in Compressed Sparse Row format>

In [219]: b.A
Out[219]: 
array([[10,  5,  0],
       [ 3,  4,  0],
       [ 0,  0,  0]])

来源：https://stackoverflow.com/questions/32800395/create-csr-matrix-from-x-index-y-index-value

标签

python

numpy

scipy

sparse-matrix