pytorch实现施密特正交化

对于⼀个给定的pytorch张量，如果想要对这⾥⾯的每⼀个⾏向量进⾏施密特正交化，可以使⽤下⾯的代码：

import numpy as np

from sympy.matrices import Matrix, GramSchmidt

import torch

functional as F

def orthogo_tensor(x):

m, n = x.size()

x_np = x.t().numpy()

matrix =[Matrix(col)for col in x_np.T]

gram = GramSchmidt(matrix)

ort_list =[]

for i in range(m):

vector =[]

for j in range(n):

vector.append(float(gram[i][j]))

ort_list.append(vector)

ort_list = np.mat(ort_list)

ort_list = torch.from_numpy(ort_list)

ort_list = F.normalize(ort_list,dim=1)

return ort_list

x = torch.randn(4,6)

x = orthogo_tensor(x)

print(x)

print(x.matmul(x.t()))

程序运⾏结果如下所⽰：

tensor([[-0.1633,0.5601,-0.6081,-0.2640,-0.4215,-0.2061],

[0.0386,-0.0236,-0.0674,0.4131,-0.5576,0.7154],

[-0.3210,0.6896,0.5430,-0.1018,0.1794,0.2898],

一叶荻[-0.3445,-0.3674,0.3962,-0.5217,-0.5563,-0.0886]],

dtype=torch.float64)

淳于髡tensor([[1.0000e+00,-3.6843e-09,1.5446e-08,-7.1455e-08],

易切削不锈钢>四川康定地震[-3.6843e-09,1.0000e+00,-2.1965e-08,-3.5087e-08],

[1.5446e-08,-2.1965e-08,1.0000e+00,-2.0042e-07],

性之初[-7.1455e-08,-3.5087e-08,-2.0042e-07,1.0000e+00]],

dtype=torch.float64)blc

本文发布于:2024-09-23 02:20:51，感谢您对本站的认可！

标签：正交康定结果

留言与评论（共有 0 条评论）