First order conditions for semidefinite representations of convex sets defined by rational or singular polynomials (Q662299): Difference between revisions

A set in \({\mathbb R}^n\) which is the projection of a linear section of the semidefinite cone \({\mathbb S}^p\) (i.e. the cone of symmetric real matrices of size \(m\) with nonnegative eigenvalues) is called semidefinite representable. Such sets are convex semialgebraic, and it is currently conjectured that all convex semialgebraic sets are semidefinite representable, provided the dimension \(p\) of the lifting cone is large enough (but finite). In this paper the author addresses the question of explicitly constructing semidefinite representations for a particular class of basic semialgebraic sets \(\{x \in {\mathcal D} :\;f(x) \geq 0\}\) with \(f(x)\) a concave polynomial or rational function, defined over a convex set \({\mathcal D} = \{x \in {\mathbb R}^n : g_1(x) \geq 0, \dots, g_m(x) \geq 0\}\) with \(g_i(x)\) are concave polynomials. As shown previously by other authors, the key condition for semidefinite representation is algebraic in nature, and can be formulated as the inclusion of a linear function of \(f\) its gradient into a preorder or quadratic module associated with the particular representation of set \(\mathcal D\) with polynomials \(g_i\). In the tricky case of semialgebraic convex sets with singularities of \(f\) along the boundary, the paper contains ad hoc, but interesting new explicit constructions of semidefinite representations.