我希望能够使用 Python 类作为元素进行矩阵运算 - 在本例中,是一个简单的伽罗瓦域 http://en.wikipedia.org/wiki/Galois_field执行。它实现了必要的__add__
, __mul__
, __sub__
etc.
起初,我认为这应该可以numpy 数组 http://docs.scipy.org/doc/numpy/reference/arrays.ndarray.html, 使用dtype
参数,但来自the dtype文档 http://docs.scipy.org/doc/numpy/reference/arrays.dtypes.html, 看起来dtype
不能是任意的 Python 类。例如,我有一堂课Galois
它进行模 2 运算:
>>> from galois import Galois
>>> Galois(1) + Galois(0)
Galois(1)
>>> Galois(1) + Galois(1)
Galois(0)
我可以尝试在 numpy 中使用它:
>>> import numpy as np
>>> a = np.identity(4, Galois)
>>> a
array([[1, 0, 0, 0],
[0, 1, 0, 0],
[0, 0, 1, 0],
[0, 0, 0, 1]], dtype=object)
但是,如果我对矩阵进行运算,则元素不遵循我的类的方法:
>>> b = np.identity(4, Galois)
>>> a+b
array([[2, 0, 0, 0],
[0, 2, 0, 0],
[0, 0, 2, 0],
[0, 0, 0, 2]], dtype=object)
有什么办法可以让这个与 numpy 一起工作吗?
是否有其他 Python 矩阵库可以对任意类似数字的类进行矩阵运算(包括求逆)?
Update
感谢到目前为止的回答。但我仍然无法像我希望的那样真正使用它。加法和乘法看起来不错,但矩阵求逆却不行。例如,让我们尝试获取AES 逆 S 盒仿射变换矩阵 http://en.wikipedia.org/wiki/Rijndael_S-box#Inverse_S-box来自前向S盒仿射变换矩阵 http://en.wikipedia.org/wiki/Rijndael_S-box#Forward_S-box.
class Galois(object):
MODULO = 2
def __init__(self, val):
self.val = int(val) % self.MODULO
def __add__(self, val):
return self.__class__((self.val + int(val)) % self.MODULO)
def __sub__(self, val):
return self.__class__((self.val - int(val)) % self.MODULO)
def __mul__(self, val):
return self.__class__((self.val * int(val)) % self.MODULO)
def __int__(self):
return self.val
def __repr__(self):
return "%s(%d)" % (self.__class__.__name__, self.val)
def __float__(self):
return float(self.val)
if __name__ == "__main__":
import numpy as np
Gv = np.vectorize(Galois)
a = Gv(np.identity(8)) + Gv(np.eye(8,8,-1)) + Gv(np.eye(8,8,-2)) + Gv(np.eye(8,8,-3)) + Gv(np.eye(8,8,-4)) + Gv(np.eye(8,8,4)) + Gv(np.eye(8,8,5)) + Gv(np.eye(8,8,6)) + Gv(np.eye(8,8,7))
print np.matrix(a)
print np.matrix(a).I
结果:
[[Galois(1) Galois(0) Galois(0) Galois(0) Galois(1) Galois(1) Galois(1)
Galois(1)]
[Galois(1) Galois(1) Galois(0) Galois(0) Galois(0) Galois(1) Galois(1)
Galois(1)]
[Galois(1) Galois(1) Galois(1) Galois(0) Galois(0) Galois(0) Galois(1)
Galois(1)]
[Galois(1) Galois(1) Galois(1) Galois(1) Galois(0) Galois(0) Galois(0)
Galois(1)]
[Galois(1) Galois(1) Galois(1) Galois(1) Galois(1) Galois(0) Galois(0)
Galois(0)]
[Galois(0) Galois(1) Galois(1) Galois(1) Galois(1) Galois(1) Galois(0)
Galois(0)]
[Galois(0) Galois(0) Galois(1) Galois(1) Galois(1) Galois(1) Galois(1)
Galois(0)]
[Galois(0) Galois(0) Galois(0) Galois(1) Galois(1) Galois(1) Galois(1)
Galois(1)]]
[[ 0.4 0.4 -0.6 0.4 0.4 -0.6 0.4 -0.6]
[-0.6 0.4 0.4 -0.6 0.4 0.4 -0.6 0.4]
[ 0.4 -0.6 0.4 0.4 -0.6 0.4 0.4 -0.6]
[-0.6 0.4 -0.6 0.4 0.4 -0.6 0.4 0.4]
[ 0.4 -0.6 0.4 -0.6 0.4 0.4 -0.6 0.4]
[ 0.4 0.4 -0.6 0.4 -0.6 0.4 0.4 -0.6]
[-0.6 0.4 0.4 -0.6 0.4 -0.6 0.4 0.4]
[ 0.4 -0.6 0.4 0.4 -0.6 0.4 -0.6 0.4]]
不是我希望的结果。看起来,对于矩阵求逆,numpy 只是将矩阵转换为浮点数,然后用普通实数进行求逆。