ITERATIVE MINIMIZATION OF QUADRATIC FUNCTIONALS

A complete solution is given to the problem of determining when a cert ain iteration will converge to the minimum of an important type of qua dratic functional. It is shown that convergence occurs whenever the mi nimum exists, and that the iterates produced by the iteration will be unbounded for every starting point if the minimum does not exist. Appl ications are given concerning an adaptive filtering algorithm and nonh armonic series expansions.