On generalizing Descartes' rule of signs to hypersurfaces (Q2172219)

Recall that the (classical) Descartes' rule of signs (for univariate polynomials) provides an upper bound on the number of positive real roots of a univariate polynomial with real coefficients in terms of the sign pattern of the coefficient vector of the polynomial. In other words, the Descartes' rule of signs provide an upper bounds on the number of connected components of $\mathbb{R}_{>0}$ minus the zero set of the univariate polynomial. There are many trials to generalize the Descartes' rule of signs in various directions. One of the efforts is to study the upper bounds on the number of solutions in the positive orthant using sign properties of the coefficients of system of $n$ polynomial equations in $n$ variables. In this paper, they introduce a new direction of the generalization of the Descartes' rule of signs to multivariate polynomials (with real exponents). Main Problem: Suppose $f:\mathbb{R}^n_{>0}\rightarrow \mathbb{R}$ is a multivariate polynomial on subsets $\sigma$ of the positive orthant with allowing real exponents. (The polynomial $f$ is called the signomial.) Then, find a (sharp) upper bound on the number of connected components of $\mathbb{R}^n_{>0}\setminus \{x\in\mathbb{R}^n_{>0}:f(x)=0\}$ based on the sign pattern of the coefficient vector of $f$ and the geometry of the subsets $\sigma$ in $\mathbb{R}^n_{>0}$. They provide bounds on the number of the connected components of the complement of a hypersurface in the positive orthant in some scenarios. In particular, they give conditions based on the geometrical configuration of the exponents and the sign of the coefficients that guarantee that the number of connected components where the polynomial attains a negative value is at most one or two. Also, their results fully cover the cases where such an upper bound provided by the univariate Descartes' rule of signs is one. They introduce the notation and known results on signomials in Section 2. Then, they provide bounds answering the Main Problem using separating hyperplanes (see Theorem 3.6, 3.8) in Section 3. They also give bounds found by providing conditions that guarantee that the signomial can be transformed into a convex function, while preserving the number of the connected components of the complement of hypersurface (see Theorem 4.6) in Section 4. Lastly, they compare the two approaches in Section 5.