sun_diffusion.linalg module

Utilities for basic linear algebra operations.

sun_diffusion.linalg.trace(mtrx, dim1=-1, dim2=-2)

Computes the trace of a batched matrix mtrx along a chosen pair of dimensions (dim1, dim2).

Warning

By default, trace does NOT preserve the dimensionality of the matrix.

Parameters:

• mtrx (Tensor, NDArray) – Matrix to be traced

• dim1 (int) – First dimension over which to trace

• dim2 (int) – Second dimension over which to trace

Return type:

Tensor | ndarray[tuple[int, ...], dtype[TypeVar(_ScalarType_co, bound= generic, covariant=True)]]

Returns:

Array with elements along the diagonal of (dim1, dim2) summed

sun_diffusion.linalg.transpose(mtrx, dim1=-1, dim2=-2)

Transposes a matrix mtrx by swapping a chosen pair of dimensions given by (dim1, dim2).

Parameters:

• mtrx (Tensor, NDArray) – Matrix to transpose

• dim1 (int) – First dimension in transpose pair

• dim2 (int) – Dimension to swap with dim1

Return type:

Tensor | ndarray[tuple[int, ...], dtype[TypeVar(_ScalarType_co, bound= generic, covariant=True)]]

Returns:

Transposed matrix

sun_diffusion.linalg.adjoint(U)

Computes the Hermitian adjoint of a complex-valued matrix U, which is defined mathematically defined as the complex conjugate transpose:

\[U^\dagger = (U^*)^\top\]

Parameters:

U (Tensor, NDArray) – Complex matrix

Return type:

Tensor | ndarray[tuple[int, ...], dtype[TypeVar(_ScalarType_co, bound= generic, covariant=True)]]

Returns:

Hermitian adjoint of U