调用 qr_solve 时无法从复数创建 mpf [英] Cannot create mpf from a complex number when calling qr_solve

问题描述

我进行了以下导入:

from sympy.matrices import Matrix as sy_matrix
import sympy.mpmath as sy_mp

像这样创建我的矩阵:

sysMat = sy_matrix([[0.0]*sz1]*sz2)       
resVec = sy_matrix([[0.0]]*sz2)

.Populate 然后用 python 复数(例如 1.0+1.0j)然后调用:

.Populate then with python complex numbers (eg 1.0+1.0j) and then call:

coeffVec = sy_mp.qr_solve(sysMat, resVec)

但是我得到以下异常:

  File "..\RatSMat.py", line 69, in solve
    self.coeffVec = sy_mp.qr_solve(self.sysMat, self.resVec)
  File "C:\Python27\lib\site-packages\sympy\mpmath\matrices\linalg.py", line 406, in qr_solve
    A, b = ctx.matrix(A, **kwargs).copy(), ctx.matrix(b, **kwargs).copy()
  File "C:\Python27\lib\site-packages\sympy\mpmath\matrices\matrices.py", line 328, in __init__
    A = self.ctx.matrix(args[0].tolist())
  File "C:\Python27\lib\site-packages\sympy\mpmath\matrices\matrices.py", line 301, in __init__
    self[i, j] = convert(a)
  File "C:\Python27\lib\site-packages\sympy\mpmath\ctx_mp_python.py", line 662, in convert
    return ctx._convert_fallback(x, strings)
  File "C:\Python27\lib\site-packages\sympy\mpmath\ctx_mp.py", line 614, in _convert_fallback
    raise TypeError("cannot create mpf from " + repr(x))
  TypeError: cannot create mpf from -2.27448396079674e-13*I

这似乎发生在 sympy 在输入之前设置矩阵时，这对我来说似乎很奇怪，因为 I 表示 sympy 复杂类型.想我可能在某处使用了错误的矩阵或输入类型.

It seems to occur when sympy is setting up the matrices before input, which seems odd to me since the I indicate a sympy complex type. Thinking I'm perhaps using the wrong matrices or input type somewhere.

推荐答案

问题是 mpmath 不知道如何将 Sympy 对象 I(虚数单位)转换成它的对象来表示复数.不过，它可以与 Python 的 j 一起使用，正如您所看到的 在源代码中.

The problem is that mpmath does not know how to convert Sympy object I (imaginary unit) into its object for representing complex numbers. It can work with Python's j, though, as you can see in the source.

需要注意的是，mpmath 是一个用于高精度浮点计算的库，它独立于 Sympy，它是一个符号计算库.不需要定义 Sympy 对象然后让 mpmath 使用它们.直接用mpmath就行了.

It should be noted that mpmath, a library for high-precision floating point computations, is independent of Sympy, a library for symbolic computations. There is no need to define Sympy objects and then ask mpmath to work with them. Just use mpmath directly.

import mpmath
mpmath.mp.dps = 50    # 50 digit arithmetic, for demonstration
A = mpmath.matrix([[2+3j, 1], [1, 1]])
b = mpmath.matrix([[1], [1]])
mpmath.qr_solve(A, b)

输出:

(matrix(
[[mpc(real='3.436455137311543569947658013827436034380299778494123e-55', imag='-1.2867118438861911327915233933248665673255859691704302e-55')],
 [mpc(real='1.0', imag='-1.1953121943036561823547019562093175737721941131313411e-54')]]), mpf('1.7055229691470039210052256620915903242996742650953353e-54'))

这篇关于调用 qr_solve 时无法从复数创建 mpf的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！