Torch和Numpy之——奇异值分解svd区别

import torch
import numpy as np

# 奇异值分解：把一个矩阵拆成3个矩阵

a = torch.tensor([[1., 2.], [3., 4.], [5., 6.]])
b = np.array([[1., 2.], [3., 4.], [5., 6.]])
 
print(torch.linalg.svd(a))
print(np.linalg.svd(b))

torch.svd 和torch.linalg.svd 有区别 ，torch.linalg.svd彩盒np.linalg.svd有一致性

torch.svd() is deprecated in favor of torch.linalg.svd() and will be removed in a future PyTorch release.

返回结果

torch.return_types.linalg_svd(
U=tensor([[-0.2298,  0.8835,  0.4082],
        [-0.5247,  0.2408, -0.8165],
        [-0.8196, -0.4019,  0.4082]]),
S=tensor([9.5255, 0.5143]),
Vh=tensor([[-0.6196, -0.7849],
        [-0.7849,  0.6196]]))
(array([[-0.2298477 ,  0.88346102,  0.40824829],
       [-0.52474482,  0.24078249, -0.81649658],
       [-0.81964194, -0.40189603,  0.40824829]]), array([9.52551809, 0.51430058]), array([[-0.61962948, -0.78489445],
       [-0.78489445,  0.61962948]]))