PyTorch torch.linalg.svd Function


Pytorch torch 参考手册PyTorch torch Reference Manual

torch.linalg.svdis a function in the PyTorch linear algebra module used to compute the singular value decomposition (SVD) of a matrix. SVD decomposes a matrix as A = U * diag(S) * V^T.

Function Definition

torch.linalg.svd(A, full_matrices=False, out=None)

Parameters:

  • A(Tensor): The input matrix.
  • full_matrices(bool, optional): If True, returns the full U and V matrices. Defaults to False.
  • out(tuple, optional): Output tuple.

Return Value:

  • tuple: Returns a tuple (U, S, Vh), where Vh is the transpose of V.

Usage Example

Example

import torch

# Create matrix
A = torch.tensor([[1.0, 2.0, 3.0],
                  [4.0, 5.0, 6.0],
                  [7.0, 8.0, 9.0],
                  [10.0, 11.0, 12.0]])

# SVD decomposition
U, S, Vh = torch.linalg.svd(A)

print("Matrix A:")
print(A)
print("nU shape:", U.shape)
print("Singular values S:", S)
print("Vh shape:", Vh.shape)
print("nVerification: U @ diag(S) @ Vh =")
print(U @ torch.diag(S) @ Vh)

The output result is:

矩阵 A:
tensor([[ 1.,  2.,  3.],
        [ 4.,  5.,  6.],
        [ 7.,  8.,  9.],
        [10., 11., 12.]])
U 形状: torch.Size([4, 4])
奇异值 S: tensor([25.4627,  1.2907,  0.0000])
Vh 形状: torch.Size([3, 3])
验证: U @ diag(S) @ Vh =
tensor([[ 1.0000,  2.0000,  3.0000],
        [ 4.0000,  5.0000,  6.0000],
        [ 7.0000,  8.0000,  9.0000],
        [10.0000, 11.0000, 12.0000]])

Pytorch torch 参考手册PyTorch torch Reference Manual

Other Extensions