Python 中矩阵逆元素求和中的虚假值

Question

我有一个矩阵，R2。我总结了每一行中非零元素的倒数。但是，如电流输出所示，会出现虚假值（第 1 个和第 5 个值）。我怎样才能避免这种情况？

R2=array([[0.00000000e+00, 1.86647223e+10, 0.00000000e+00, 6.54826235e+09,
        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
       [1.11453512e+10, 0.00000000e+00, 2.82017381e+11, 0.00000000e+00,
        3.64932878e+10, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
       [0.00000000e+00, 1.86647223e+10, 0.00000000e+00, 0.00000000e+00,
        0.00000000e+00, 4.48759847e+09, 0.00000000e+00, 0.00000000e+00,
        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
       [1.11453512e+10, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
        3.64932878e+10, 0.00000000e+00, 1.87125503e+09, 0.00000000e+00,
        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
       [0.00000000e+00, 1.86647223e+10, 0.00000000e+00, 6.54826235e+09,
        0.00000000e+00, 4.48759847e+09, 0.00000000e+00, 4.08814184e+10,
        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
       [0.00000000e+00, 0.00000000e+00, 2.82017381e+11, 0.00000000e+00,
        3.64932878e+10, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
        1.65770709e+11, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
       [0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 6.54826235e+09,
        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 4.08814184e+10,
        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
       [0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
        3.64932878e+10, 0.00000000e+00, 1.87125503e+09, 0.00000000e+00,
        1.65770709e+11, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
       [0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
        0.00000000e+00, 4.48759847e+09, 0.00000000e+00, 4.08814184e+10,
        0.00000000e+00, 1.20130268e+10, 3.53068596e+10, 0.00000000e+00],
       [0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
        1.65770709e+11, 0.00000000e+00, 0.00000000e+00, 5.22617991e+08],
       [0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
        1.65770709e+11, 0.00000000e+00, 0.00000000e+00, 5.22617991e+08],
       [0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
        0.00000000e+00, 1.20130268e+10, 3.53068596e+10, 0.00000000e+00]])

R3=np.divide(1, R2, where=R2 != 0).sum(1)

当前输出为

array([8.94213159e+130, 1.20671688e-010, 2.76413345e-010, 6.51526508e-010,
       7.71108578e+140, 3.69806090e-011, 1.77173260e-010, 5.67835429e-010,
       3.58863402e-010, 1.91947593e-009, 1.91947593e-009, 5.22617991e+008])

Answer 1

为此，您可以通过两种方式做到这一点，一种是在您的 np.devide 中使用 out=R2 作为：

np.divide(1, R2, where=R2!=0, out=R2).sum(1)

# R2:
# [2.0628927806128987e-10 1.2067168750757737e-10 2.7641334517657591e-10
#  6.5152650788289563e-10 4.5358660476653588e-10 3.6980608990547905e-11
#  1.7717325958995994e-10 5.6783542936586616e-10 3.5886340231599305e-10
#  1.9194759328959448e-09 1.9194759328959448e-09 1.1156607561074707e-10]

或将 np.devide 结果放入另一个名为 R3 的变量中，然后将其求和为：

R3 = np.divide(1, R2, where=R2!=0)
result = R3.sum(1)

Python 中矩阵逆元素求和中的虚假值

Spurious values in summation of inverse elements of a matrix in Python

python

numpy