tf.linalg.sqrtm

Computes the matrix square root of one or more square matrices:

tf.linalg.sqrtm(
    input: Annotated[Any, TV_MatrixSquareRoot_T], name=None
) -> Annotated[Any, TV_MatrixSquareRoot_T]

matmul(sqrtm(A), sqrtm(A)) = A

The input matrix should be invertible. If the input matrix is real, it should have no eigenvalues which are real and negative (pairs of complex conjugate eigenvalues are allowed).

The matrix square root is computed by first reducing the matrix to quasi-triangular form with the real Schur decomposition. The square root of the quasi-triangular matrix is then computed directly. Details of the algorithm can be found in: Nicholas J. Higham, "Computing real square roots of a real matrix", Linear Algebra Appl., 1987.

The input is a tensor of shape [..., M, M] whose inner-most 2 dimensions form square matrices. The output is a tensor of the same shape as the input containing the matrix square root for all input submatrices [..., :, :].