mindspore.mint.linalg.qr

View Source On Gitee
mindspore.mint.linalg.qr(A, mode='reduced')

Orthogonal decomposition of the input \(A = QR\).

Where A is an input tensor, a dimension is at least 2, and A may be represented as a product form of an orthogonal matrix Q and an upper triangular matrix R.

Warning

This is an experimental API that is subject to change or deletion.

Parameters

  • A (Tensor) – The calculated matrix, A is at least two-dimensional.

  • mode (str, optional) –

    Matrix decomposition mode. The options are reduced, complete, and r. The default value is reduced.

    • "reduced": For input \(A(*, m, n)\) output simplified size \(Q(*, m, k)\), \(R(*, k, n)\), where k is the minimum value of m and n.

    • "complete": For input \(A(*, m, n)\) output full-size \(Q(*, m, m)\), \(R(*, m, n)\).

    • "r": Only \(R(*, k, n)\) in the reduced scenario is calculated, where k is the minimum value of m and n, and Q is returned as an empty tensor.

Returns

  • Q (Tensor) - The shape is \(Q(*, m, k)\) or \((*, m, n)\), has the same dtype as A.

  • R (Tensor) - The shape is \(Q(*, k, n)\) or \((*, m, n)\), has the same dtype as A.

Raises

  • TypeError – If A is not a Tensor.

  • TypeError – If the dtype of A is not the float32.

  • ValueError – If A is not empty and its dimension is less than 2 dimensions.

Supported Platforms:

Ascend

Examples

>>> import mindspore
>>> import numpy as np
>>> from mindspore import Tensor, mint
>>> x = Tensor(np.array([[1.0, 1.0, 2.0, 4.0], [1.0, 1.0, 2.0, 4.0]]), mindspore.float32)
>>> output = mint.linalg.qr(x)
>>> print(output)
(Tensor(shape=[2, 2], dtype=Float32, value=
[[-7.07106829e-01, -7.07106769e-01],
[-7.07106769e-01,  7.07106829e-01]]),
Tensor(shape=[2, 4], dtype=Float32, value=
[[-1.41421354e+00, -1.41421354e+00, -2.82842731e+00, -5.65685463e+00],
[ 0.00000000e+00,  3.42285418e-08,  0.00000000e+00,  0.00000000e+00]]))