Orthogonal polynomials, Hankel matrices, and the Lanczos algorithm

We explore the application of the nonsymmetric Lanczos algorithm to two different problem domains, the theory of moments and orthogonal polynomials, and the factorization of Hankel matrices. The connection with a third problem domain, algorithm-based fault tolerant computing, was explored in a companion paper. We find that in the simplest case, where all leading submatrices are nonsingular, the methods reduce to classical algorithms such as the original nonsymmetric Lanczos method and the Chebyshev algorithm. We propose a back-up pivoting strategy for factorizing a Hankel matrix which avoids treating rank deficiency as a special case.