I'm one of the many who is far more interested in the reason for the
request than in the numerical method to use. The way I see it, the
methodology of, say, polyroot() suffices in a mathematical sense. If
your coefficients are mathematically exact (as opposed to measured
values), the near-singularities and such don't really matter, assuming
infinite precision math. So, what I'd like to know is:
--Is this some attempt to fit horrific data to a polynomial?
--Is this the start of an attempt at finding some intermediate theorems
to simplify the root search (sort of like the theorems which allow one
to sort primes without dividing the candidates by every single lesser
prime)?
Carl
