Embedding polynomial systems into vertically parametrised families: A case study on ODEbase
Oliver Daisey, Yue Ren, Yuvraj Singh · 2025
Read Embedding polynomial systems into vertically parametrised families: A case study on ODEbase by Oliver Daisey, Yue Ren, Yuvraj Singh (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-00156.txt.
Introduction
Oliver Daisey, Yue Ren, and Yuvraj Singh study how to dress a system of polynomial equations in a well-behaved family. Vertically parametrised systems are a class whose combinatorics carries rich algebraic information, making them prized in computational algebra. Given one fixed system, what makes a good such family that contains it, and how would you build one? Their paper, published in 2024, attacks the question empirically.
For raw material the team took every polynomial system in ODEbase, a database of differential-equation models. They transcribed them into a form the OSCAR algebra system can read, released as a Julia package. Across that corpus they test what good embeddings look like and how to construct them. The work is a craftsman's study: less a single theorem than a survey of what actually works when theory meets a database of real systems.
Catalog record
| CREATOR | Oliver Daisey, Yue Ren, Yuvraj Singh |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 26 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-00156
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Embedding polynomial systems into vertically parametrised families: A case study on ODEbase Oliver Daisey, Yue Ren, Yuvraj Singh Abstract . Vertically parametrised polynomial systems are a particular nice class of parametrised polynomial systems for which a lot of interesting algebraic information is encoded in its combinatorics. Given a fixed polynomial system, we empirically study what constitutes a good vertically parametrised polynomial system that gives rise to it and how to construct said vertically parametrised polynomial system. For data, we use all polynomial systems in ODEbase , which we have transcribed to an OSCAR readable format, and made available as a Julia package OscarODEbase . 1. Introduction Vertically parametrised systems are an important class of parametrised polynomial systems. They describe the steady states in mass-action kinetics, and many of their interesting properties are encoded in their (tropical) combinatorics: (1) the solution set always has the expected dimension and at least one smooth point [ FHP24 ] , (2) if generically zero-dimensional, both (a) the generic number of complex solutions [ HR22 ] and, (b) a lower bound on the number of positive solutions for a particular instance [ RT24 ] can be computed combinatorially using tropical geometry, (3) if generically zero-dimensional, optimal homotopies can be constructed combinatorially using tropical geometry [ HHR24 ] . In this article, given a polynomial system F=\{f_{1},\dots,f_{k}\}\subseteq K[x^{\pm}] , we study the task of finding a “good” embedding of F into a vertically parametrised system \mathcal{F} . Specifically, our goal is to find a vertically parametrised system \mathcal{F} such that F=\mathcal{F}_{P} for some choice of parameters P\in K^{m} with \mathcal{F}_{P} denoting the specialisation of \mathcal{F} at P . By “good”, we mean that the system F exhibits many generic properties of \mathcal{F} . For example, the number of solutions of F should coincide with the generic root count of \mathcal{F} . This naturally leads to the first problem: Problem 1: Identify easy criteria for good embeddings. In Section 3 , we turn to the polynomial networks in ODEbase [ LSR22 ] . We consider a random specialisation of each vertically parametrised system and identify which features distinguish the original parametrised system from other embeddings of its specialisation. We find that (a) the original parametrised system minimises the number of distinct monomials, (b) there is no conclusive tiebreaker among all perturbations minimising the number of distinct monomials. This then naturally leads to the second problem Problem 2: Construct a good embedding. In Section 4 , we show that constructing an optimal embedding is an incredibly challenging task. Instead, we propose a greedy algorithm and compare its performance to the optimum. In Section 5 , we introduce our OSCAR [ OSCAR ] interface to the ODEbase database and explain how it can be used. 2. Background In this section, we go over some basic notions that are of immediate interest. 2.1. Parametrised polynomial systems We begin with introducing some notation for parametrised polynomials. Our notation will be the same as in [ HHR24 ] . Convention 2.1 For the entirety of the paper, we fix a parametrised (Laurent) polynomial ring K[a][x^{\pm}]\coloneqq K[a_{1},\dots,a_{m}][x_{1}^{\pm},\dots,x_{n}^{\pm}] with parameters a_{1},\dots,a_{m} and variables x_{1},\dots,x_{n} . Elements \dutchcal{f}\in K[a][x^{\pm}] are referred to as parametrised polynomials , and ideals \mathcal{I}\subseteq K[a][x^{\pm}] are referred to as parametrised polynomial ideals . Finite \mathcal{F}=\{\dutchcal{f}_{1},\dots,\dutchcal{f}_{k}\}\subseteq K[a][x^{\pm}] are parametrised polynomial systems . Points P=(P_{1},\dots,P_{m})\in K^{m} are called choices of parameters and we use subscripts to denote specialisations thereover: \displaystyle\dutchcal{f}_{P} \displaystyle\coloneqq\sum_{\alpha\in S}c_{\alpha}(P)\cdot x^{\alpha}\in K[x^{\pm}], provided \dutchcal{f}=\sum_{\alpha\in S}c_{\alpha}\cdot x^{\alpha} with c_{\alpha}\in K[a] and S\subseteq\mathbb{Z}^{n} finite, and \displaystyle\mathcal{I}_{P} \displaystyle\coloneqq\langle\dutchcal{f}_{P}\mid\dutchcal{f}\in\mathcal{I}\rangle\subseteq K[x^{\pm}], \displaystyle\mathcal{F}_{P} \displaystyle\coloneqq\{\dutchcal{f}_{1,P},\dots,\dutchcal{f}_{k,P}\}\subseteq K[x^{\pm}]. Definition 2.2 The generic specialisation of a parametrised polynomial \dutchcal{f}=\sum_{\alpha\in S}c_{\alpha}x^{\alpha}\in K[a][x^{\pm}] is the polynomial over the rational function field K(a)\coloneqq K(a_{1},\dots,a_{m}) given by \dutchcal{f}_{K(a)}\coloneqq\sum_{\alpha\in S}c_{\alpha}x^{\alpha}\in K(a)[x^{\pm}]. The generic specialisation of a parametrised polynomial ideal \mathcal{I}\subseteq K[a][x^{\pm}] is the polynomial ideal over K(a) given by \mathcal{I}_{K(a)}\coloneqq\langle\dutchcal{f}_{K(a)}\mid\dutchcal{f}\in\mathcal{I}\rangle\subseteq K(a)[x^{\pm}]. Definition 2.3 Let \mathcal{I}\subseteq K[a][x^{\pm}] be a parametrised polynomial ideal. The generic dimension of \mathcal{I} is the Krull dimension of its generic specialisation \mathcal{I}_{K(a)} . The generic root count of I is \ell_{\mathcal{I},K(a)}\coloneqq\infty if \dim(\mathcal{I}_{K(a)})>0 or it is the vector space dimension \ell_{\mathcal{I},K(a)}\coloneqq\dim_{K(a)}{\raisebox{1.99997pt}{$K[x^{\pm}]$}\left/\raisebox{-1.99997pt}{$\mathcal{I}_{K(a)}$}\right.}. The generic degree of \mathcal{I} is the generic root count of \mathcal{I}+I_{L} , where I_{L}\subseteq K[x^{\pm}] is a generic linear ideal with \codim(I_{L})=\dim(\mathcal{I}_{K(a)}) . The generic root count can be regarded as the number of solutions of \mathcal{I}_{P} for a generic choice of parameters P\in K^{m} and counted with a suitable multiplicity [ CLO05 , Corollary 2.5] . Consequently the generic degree counts the number of intersection points of V(\mathcal{I}_{P})\cap L for a generic choice of parameters P\in K^{m} and with a generic linear space L\subseteq(K^{\ast})^{n} of complementary dimension. 2.2. Vertically parametrised ideals Next we introduce vertically parametrised ideals and embeddings of polynomial systems into them. We recall a central notion of [ FHP24 ] , which implies that the (generic) dimension is independent of the embedding. Definition 2.4 A vertically parametrised polynomial system is a system \mathcal{F}=\{\dutchcal{f}_{1},\dots,\dutchcal{f}_{k}\}\subseteq K[a][x^{\pm}] of the form (1) \dutchcal{f}_{i}\coloneqq\sum_{j=1}^{m}c_{i,j}a_{j}x^{\alpha_{j}} for some finite set S=\{\alpha_{1},\dots,\alpha_{m}\}\subseteq\mathbb{Z}^{n} and some c_{i,j}\in K . We generally assume the \alpha_{j} to be pairwise distinct, k\leq n\leq m , and that for all \alpha_{j}\in S there is an i\in[k] with c_{i,j}\neq 0 . We will refer to the ideal \mathcal{I}\coloneqq\langle\mathcal{F}\rangle\subseteq K[a][x^{\pm}] as a vertically parametrised ideal. Definition 2.5 Let F=\{f_{1},\dots,f_{k}\}\subseteq K[x^{\pm}] be a polynomial system, say f_{i}=\sum_{j=1}^{m}c_{i,j}x^{\alpha_{j}} for some S=\{\alpha_{1},\dots,\alpha_{m}\}\subseteq\mathbb{Z}^{n} , and let I\coloneqq\langle F\rangle\subseteq K[x^{\pm}] be the ideal it generates. The minimal vertical system specialising to F is the parametrised system \mathcal{F} from Equation 1 . We denote its parametrised polynomial ideal by \mathcal{I}_{F}\coloneqq\langle\mathcal{F}\rangle , it naturally specialises to I . We will refer to \mathcal{I} as a vertical family containing I , as it can be regarded as a set of polynomial ideals of which one is I . Example 2.6 Consider the polynomial ideal I\coloneqq\langle\underbrace{x_{1}^{2}+x_{2}^{2}+x_{1},x_{1}^{2}+x_{2}^{2}+1}_{\eqqcolon F}\rangle=\langle\underbrace{x_{1}^{2}+x_{2}^{2}+x_{1},x_{1}^{2}x_{2}+x_{2}^{3}+x_{2}}_{\eqqcolon G}\rangle\subseteq\mathbb{C}[x_{1}^{\pm},x_{2}^{\pm}], and let F,G\subseteq\mathbb{C}[x^{\pm}] be the two polynomial systems above. The minimal parametrised systems specialising to F and G are \mathcal{F}\coloneqq\left\{\begin{array}[]{l}a_{1}x_{1}^{2}+a_{2}x_{2}^{2}+a_{3}x_{1}\\ a_{1}x_{1}^{2}+a_{2}x_{2}^{2}+a_{4}\end{array}\right\}\qquad\text{and}\qquad\mathcal{G}\coloneqq\left\{\begin{array}[]{l}b_{1}x_{1}^{2}+b_{2}x_{2}^{2}+b_{3}x_{1}\\ b_{4}x_{1}^{2}x_{2}+b_{5}x_{2}^{3}+b_{6}x_{2}\end{array}\right\}. They give rise to two distinct vertically parametrised ideals \mathcal{I}_{F}\coloneqq\langle\mathcal{F}\rangle\subseteq K[a][x^{\pm}] and \mathcal{I}_{G}\coloneqq\langle\mathcal{G}\rangle\subseteq K[b][x^{\pm}] in distinct parametrised polynomial rings K[a][x^{\pm}]\coloneqq K[a_{1},\dots,a_{4}][x^{\pm}] and K[b][x^{\pm}]\coloneqq K[b_{1},\dots,b_{6}][x^{\pm}] . It is straightforward to see that \ell_{I}=\ell_{\mathcal{I}_{F},K(a)}=2<4=\ell_{\mathcal{I}_{G},K(b)}. The reason why the generic root count of \mathcal{I}_{F} is lower than that of \mathcal{I}_{G} can be attributed to cancellations in the higher degree monomials of \mathcal{F} . We close this subsection with two results from existing literature on minimal vertical systems. Lemma 2.7 shows that the generic dimension of the minimal vertical system is independent on the choice of system. Lemma 2.8 shows that the generic degree of the minimal vertical system is increasing in the monomial support. Lemma 2.7 Let I\subseteq K[x^{\pm}] be a complete intersection of codimension r , and let F=\{f_{1},\dots,f_{r}\} be any generating set of I . Let \mathcal{I}_{F}\subseteq K[a][x^{\pm}] be the vertically parametrised ideal induced by F , and let \mathcal{I}_{F,K(a)} denote the generic specialisation. Then \dim(\mathcal{I}_{F,K(a)})=n-r . Proof. By [ FHP24a , Theorem 3.1] , the incidence variety \Big\{(P,x)\in K^{m}\times(K^{*})^{n}\mid x\in V(\mathcal{I}_{F,P})\Big\} is irreducible of dimension m+n-r . Thus by [ FHP24a , Theorem 2.2] , the generic specialisation \mathcal{I}_{F,K(a)} has dimension n-r . ∎ Lemma 2.8 Let I\subseteq K[x^{\pm}] be a complete intersection of codimension r , and let F=\{f_{1},\dots,f_{r}\},G=\{g_{1},\dots,g_{r}\} be