The classical Jacobi algorithm is extended to an unified Lie algebraic approach. The conventional Jacobi algorithm minimize the distance to diagonality; they reduce the off-norm, i. e. the sum of squares of off-diagonal entries. Sorting the diagonal elements after each step would accelerate the convergence but, there are difficulties to apply this sorting to the off-norm, that has to minimize. Using the gradient flow of a trace function [it R. W. Brockett, ``Dynamical systems that sort lists, diagonalize matrices, and solve linear programming problems", Linear Algebra Appl. 146, 79--91 (1991; Zbl 0719.90045)] as a more appropriate distance measure matrices can be diagonalized, and the eigenvalues can be simultaneously sorted. In this paper the sort-Jacobi algorithm is extended to a large class of structured matrices, besides the semisimple Lie algebra cases, such as e. g. the symmetric, Hermitian, and skew-symmetric eigenvalue problem, and the real and complex singular value decomposition. par The local quadratic convergence of the sort-Jacobi method is proved for the regular case but, not for the case in which the eigenvalues or singular values occur in clusters. For these more complicated irregular problem a future publication is announced. par The results are illustrated considering the Lie algebra of derivations of the complex octonions.     «

The classical Jacobi algorithm is extended to an unified Lie algebraic approach. The conventional Jacobi algorithm minimize the distance to diagonality; they reduce the off-norm, i. e. the sum of squares of off-diagonal entries. Sorting the diagonal elements after each step would accelerate the convergence but, there are difficulties to apply this sorting to the off-norm, that has to minimize. Using the gradient flow of a trace function [it R. W. Brockett, ``Dynamical systems that sort lists, d...     »