Parametrized systems of polynomial inequalitites with real exponents via linear algebra and convex geometry

Stefan Müller, Georg Regensburger
Mathematics, Algebraic Geometry, Algebraic Geometry (math.AG), Molecular Networks (q-bio.MN)
2023-06-23 16:00:00
We consider parametrized systems of generalized polynomial inequalities (with real exponents) in $n$ positive variables $x$ and involving $m$ monomial terms $c \circ x^B$ (with exponent matrix $B$, positive parameters $c$, and component-wise product $\circ$). More generally, we study solutions $x \in \mathbb{R}^n_>$ to $(c \circ x^B) \in C$ for a "coefficient cone" $C \subseteq \mathbb{R}^m_>$. Our framework encompasses systems of generalized polynomial equations, as studied in real fewnomial and reaction network theory. We identify the relevant geometric objects of the problem, namely a bounded set $P$ arising from the coefficient cone and two subspaces representing monomial differences and dependencies. The dimension of the latter subspace (the monomial dependency $d$) is crucial. As our main result, we rewrite the problem in terms of $d$ binomial conditions on the "coefficient set" $P$, involving $d$ monomials in the parameters. In particular, we obtain a classification of polynomial inequality systems. If $d=0$, solutions exist (for all $c$) and can be parametrized explicitly, thereby generalizing monomial parametrizations (of the solutions). Even if $d>0$, solutions on the coefficient set can often be determined more easily than solutions of the original system. Our framework allows a unified treatment of multivariate polynomial inequalities, based on linear algebra and convex geometry. We demonstrate its novelty and relevance in three examples from real fewnomial and reaction network theory. For two mass-action systems, we parametrize the set of equilibria and the region for multistationarity, respectively, and for univariate trinomials, we provide a "solution formula" involving discriminants and "roots".
PDF: Parametrized systems of polynomial inequalitites with real exponents via linear algebra and convex geometry.pdf
Empowered by ChatGPT