Compound matrices: Properties, numerical issues and analytical computations (Q1014362)

The authors study the problem, how to calculate efficiently compounds of real matrices that have a special form or structure. The usefulness of such a study lies in the fact that the computation of compound matrices, which is generally noneffective due to its high complexity, is encountered in several applications. A new approach to compute the singular value decompositions (SVDs) of the compounds of a matrix is proposed by establishing the equality (up to a permutation) between the compounds of the SVD of a matrix and the SVD's of the compounds of the matrix. The superiority of the new approach over the standard method is demonstrated. Similar techniques with some limitations can be adopted for other matrix factorizations, too. Furthermore, formulas for the \(n - 1\) compounds of Hadamard matrices are derived, which dodge the strenuous computations of the respective numerous large determinants. Finally, a combinatorial counting technique for finding the compounds of diagonal matrices is illustrated. Several numerical examples are given.