General non-realizability certificates for spheres with linear programming. (January 2023)

• Record Type: Journal Article
• Authors: Gouveia, João; Macchia, Antonio; Wiebe, Amy
• Is Part Of: Journal of symbolic computation. Volume 114(2023)
• Page Start: 172
• Page End: 192
• Publication Date: 2023-01
• Subjects: Non-realizability certificates -- Final polynomials -- Slack matrices -- Linear programming; Mathematics -- Data processing -- Periodicals; Numerical analysis -- Data processing -- Periodicals; Automatic programming (Computer science) -- Periodicals; Mathématiques -- Informatique -- Périodiques; Analyse numérique -- Informatique -- Périodiques; Programmation automatique -- Périodiques; Automatic programming (Computer science); Mathematics -- Data processing; Numerical analysis -- Data processing; Periodicals; Electronic journals; 510.285
• DOI: 10.1016/j.jsc.2022.04.013
• Languages: English
• ISSNs: 0747-7171

Abstract

Abstract: In this paper we present a simple technique to derive certificates of non-realizability for a combinatorial polytope. Our approach uses a variant of the classical algebraic certificates introduced by Bokowski and Sturmfels (1989), the final polynomials. More specifically we reduce the problem of finding a realization to that of finding a positive point in a variety and try to find a polynomial with positive coefficients in the generating ideal (a positive polynomial), showing that such point does not exist. Many, if not most, of the techniques for proving non-realizability developed in the last three decades can be seen as following this framework, using more or less elaborate ways of constructing such positive polynomials. Our proposal is more straightforward as we simply use linear programming to exhaustively search for such positive polynomials in the ideal restricted to some linear subspace. Somewhat surprisingly, this elementary strategy yields results that are competitive with more elaborate alternatives, and allows us to derive new examples of non-realizable combinatorial polytopes.