Signal flow graph approach to inversion of \((H,m)\)-quasiseparable-Vandermonde matrices and new filter structures (Q848585): Difference between revisions

The language of signal flow graphs, common in electrical engineering, control theory, etc., provides also a tool for proving purely mathematical results for structured matrices using (preferably sparse) recurrence relations of the corresponding polynomial systems. The authors focus on \((H,m)\)-quasiseparable and \((H,m)\)-semiseparable matrices and related polynomials (definitions are given). Three important mathematical problems are solved using the language of signal flow graphs: efficient polynomial evaluation (monomials, Horner polynomials as examples of the \((H,1)\)-quasiseparable and \((H,1)\)-semiseparable polynomials; real-orthogonal polynomials; Szegö polynomials), computing eigenvectors of \((H,m)\)-quasiseparable and \((H,m)\)-semiseparable matrices if their eigenvalues are given, and fast inversion of \((H,m)\)-quasiseparable-Vandermonde and \((H,m)\)-semiseparable-Vandermonde matrices. The building blocks of signal flow graphs are given and explained. The particular graphs, that realize the proofs of the mathematical theorems, are much more complicated. The authors give these proofs using the language of the signal flow graphs and cite papers in which algebraic proofs can be found. Two new filter classes corresponding to both \((H,m)\)-quasiseparable matrices and their subclass \((H,m)\)-semiseparable matrices are given and classified in terms of recurrent relations.