Given a real δ, find stable real polynomials p and q such that the polynomial r(s) = (s − 2δs + 1)p(s) + (s − 1)q(s) is also stable. (We call a polynomial p stable if its abscissa α(p) = max{Re s : p(s) = 0} is nonpositive.) Clearly the problem is unsolvable if δ = 1, since then r(1) = 0; more delicate results (summarized in [7]) show it remains unsolvable for δ < 1 close to 1. Blondel offered a prize of 1kg of Belgian chocolate for the case δ = 0.9, a problem solved via randomized search in [7]. To illustrate the theme of this talk, we first outline (based on joint work with D. Henrion) a more systematic, optimization approach to the chocolate problem. We fix the degrees of the polynomials p and q (say 3, for example), without loss of generality suppose p is monic, and consider the resulting problem
[1]
Adrian S. Lewis,et al.
Active Sets, Nonsmoothness, and Sensitivity
,
2002,
SIAM J. Optim..
[2]
Adrian S. Lewis,et al.
A Robust Gradient Sampling Algorithm for Nonsmooth, Nonconvex Optimization
,
2005,
SIAM J. Optim..
[3]
Vijay V. Patel,et al.
Some applications of randomized algorithms for control system design
,
2002,
Autom..
[4]
Michael L. Overton,et al.
Variational Analysis of the Abscissa Mapping for Polynomials
,
2000,
SIAM J. Control. Optim..
[5]
Adrian S. Lewis,et al.
Approximating Subdifferentials by Random Sampling of Gradients
,
2002,
Math. Oper. Res..