Octave与Python中复数算术运算结果不一致问题求助

现象说明 #

单独计算v和w时，Octave与Python的输出结果非常接近：

Octave中v的输出 #

v octave values =

 Columns 1 through 3:

          0 - 223.4950i          0 - 194.4942i          0 - 167.4933i

 ...

 Columns 16 through 18:

          0 +        0i     1.4142 +        0i          0 +   2.0000i

 ... 
 Columns 28 through 30:

          0 + 142.4921i          0 + 167.4933i          0 + 194.4942i

 Column 31:

          0 + 223.4950i

Python中v的输出 #

v python values 
 [-0.        -223.49496639j -0.        -194.49421585j
 -0.        -167.49328345j -0.        -142.49210504j
...
 -1.41421356  +0.j          0.          +0.j
 ...
  0.        +167.49328345j  0.        +194.49421585j
  0.        +223.49496639j]

Octave中w的输出 #

w octave values =

Columns 1 through 4: %before 0

  -4.7124 + 8.5456i  -4.7124 + 8.3371i  -4.7124 + 8.1130i  -4.7124 + 7.8705i

 ...

 Columns 13 through 16:

  -4.7124 + 3.4386i  -4.7124 + 1.6479i  -1.8464 +      0i        0 +      0i

...

 Columns 25 through 28: %after 0

   4.7124 - 6.9950i   4.7124 - 7.3165i   4.7124 - 7.6064i   4.7124 - 7.8705i

 Columns 29 through 31:

   4.7124 - 8.1130i   4.7124 - 8.3371i   4.7124 - 8.5456i

Python中w的输出 #

w python values 
 [-4.71238898+8.54562334j -4.71238898+8.33714905j -4.71238898+8.11296806j
 ...
  0.        +0.j          1.84643913+0.j          
...
  4.71238898+7.87049456j  4.71238898+8.11296806j  4.71238898+8.33714905j
  4.71238898+8.54562334j]

但计算z = v + w时，两者结果差异明显：

Octave中z的输出 #

z = v+w octave values = 

 Columns 1 through 3:

    -4.7124 - 214.9493i    -4.7124 - 186.1571i    -4.7124 - 159.3803i
 ...
 Columns 16 through 18: %after 0

          0 +        0i     3.2607 +        0i     4.7124 +   0.3521i

 Columns 19 through 21:

     4.7124 +   3.9098i     4.7124 +   9.9835i     4.7124 +  18.2866i

 ...
 Columns 28 through 30:

     4.7124 + 134.6216i     4.7124 + 159.3803i     4.7124 + 186.1571i

 Column 31:

     4.7124 + 214.9493i

Python中z的输出 #

z= v+w python values 
 [-4.71238898-214.94934305j -4.71238898-186.1570668j
 -4.71238898-159.38031539j -4.71238898-134.62161049j
 ...
-3.26065269  +0.j          0.          +0.j
  3.26065269  +0.j          4.71238898  +3.64791843j
 ...
  4.71238898+127.09702398j  4.71238898+150.3625996j
  4.71238898+175.60625151j  4.71238898+202.8313649j
  4.71238898+232.04058972j]

代码实现 #

Octave代码 #

%octave code

x = -15:1:15;
E = 3;

p = sqrt(E - x.^2);
v = x.*p;

w = E.*asin(x./sqrt(E));

z = w + v;

Python代码 #

#python Code
import numpy as np

x = np.arange(-15, 16, 1)
E = 3

p = np.sqrt(E - x.astype(complex)**2)
v = x.astype(complex)*p

w = E*np.arcsin(x.astype(complex)/np.sqrt(E))

z = w + v

问题 #

请问该差异产生的原因是什么？如何解决这个问题？

内容的提问来源于stack exchange，提问作者user20186251