python - scipy 稀疏矩阵除法

Question

我一直在尝试将 python scipy 稀疏矩阵除以其行的向量和。这是我的代码

sparse_mat = bsr_matrix((l_data, (l_row, l_col)), dtype=float)
sparse_mat = sparse_mat / (sparse_mat.sum(axis = 1)[:,None])

然而，无论我如何尝试，它都会抛出错误

sparse_mat = sparse_mat / (sparse_mat.sum(axis = 1)[:,None])
File "/usr/lib/python2.7/dist-packages/scipy/sparse/base.py", line 381, in __div__
return self.__truediv__(other)
File "/usr/lib/python2.7/dist-packages/scipy/sparse/compressed.py", line 427, in __truediv__
raise NotImplementedError
NotImplementedError

有人知道我哪里出错了吗？

Answer 1

您可以通过从行总和的倒数创建一个稀疏对角矩阵，然后将其与您的矩阵相乘来规避这个问题。在产品中，对角矩阵向左，您的矩阵向右。

例子:

>>> a
array([[0, 9, 0, 0, 1, 0],
       [2, 0, 5, 0, 0, 9],
       [0, 2, 0, 0, 0, 0],
       [2, 0, 0, 0, 0, 0],
       [0, 9, 5, 3, 0, 7],
       [1, 0, 0, 8, 9, 0]])
>>> b = sparse.bsr_matrix(a)
>>> 
>>> c = sparse.diags(1/b.sum(axis=1).A.ravel())
>>> # on older scipy versions the offsets parameter (default 0)
... # is a required argument, thus
... # c = sparse.diags(1/b.sum(axis=1).A.ravel(), 0)
...
>>> a/a.sum(axis=1, keepdims=True)
array([[ 0.        ,  0.9       ,  0.        ,  0.        ,  0.1       ,  0.        ],
       [ 0.125     ,  0.        ,  0.3125    ,  0.        ,  0.        ,  0.5625    ],
       [ 0.        ,  1.        ,  0.        ,  0.        ,  0.        ,  0.        ],
       [ 1.        ,  0.        ,  0.        ,  0.        ,  0.        ,  0.        ],
       [ 0.        ,  0.375     ,  0.20833333,  0.125     ,  0.        ,  0.29166667],
       [ 0.05555556,  0.        ,  0.        ,  0.44444444,  0.5       ,  0.        ]])
>>> (c @ b).todense() # on Python < 3.5 replace c @ b with c.dot(b)
matrix([[ 0.        ,  0.9       ,  0.        ,  0.        ,  0.1       ,  0.        ],
        [ 0.125     ,  0.        ,  0.3125    ,  0.        ,  0.        ,  0.5625    ],
        [ 0.        ,  1.        ,  0.        ,  0.        ,  0.        ,  0.        ],
        [ 1.        ,  0.        ,  0.        ,  0.        ,  0.        ,  0.        ],
        [ 0.        ,  0.375     ,  0.20833333,  0.125     ,  0.        ,  0.29166667],
        [ 0.05555556,  0.        ,  0.        ,  0.44444444,  0.5       ,  0.        ]])