Is there any way to tell how many clusters there are with respect to all the roots of a polynomial?

Specifically, I'm after the multiplicity of each root but since I would like to work in floating-point arithmetic I'm afraid I have to deal with clusters.

I don't mind any method of finding out: be it by numerical-iterative means during the convergence, a priori/a posteriori guess, maybe some matrix method would help..? If I have to set some small disk radius, that's ok, too.

  • $\begingroup$ Just so I understand - By cluster you mean several roots of the polynomial that are very close to each other (or the same root repeated several times) right? $\endgroup$ – Oria Gruber Sep 16 '15 at 23:36
  • $\begingroup$ @OriaGruber Yes. $\endgroup$ – Ecir Hana Sep 16 '15 at 23:38

Probably the single most important result you need to know is Rouché's theorem.

However, for postprocessing the output of numerical iterative methods (Aberth-Ehrlich, Durand-Kerner, ...), have a look at Siegfried Rump's Ten methods to bound multiple roots of polynomials; in particular methods 4 and 5 by Neumaier (based on Gershgorin circles of the companion matrix in a suitable basis).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.