Determinantal computation of minimal local GADs

Oriol Reig Fité, Daniele Taufer · 2026

Read Determinantal computation of minimal local GADs by Oriol Reig Fité, Daniele Taufer (2026) 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-2603-08836.txt.

Introduction

Oriol Reig Fité and Daniele Taufer compute minimal local generalized additive decompositions of homogeneous polynomials. They check the build does not depend on the chosen apolarity action, then give a determinant-based method that cuts the rank of a symbolic inverse system. If minimal supports are finite, the method lists every minimal decomposition without tensor extensions, and this holds when the rank stays within the degree.

Catalog record

CREATOROriol Reig Fité, Daniele Taufer
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE68558
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.08836

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Determinantal computation of minimal local GADs Oriol Reig Fité, Daniele Taufer Abstract . We study local generalized additive decompositions (GADs) of homogeneous polynomials and their associated point schemes through their local inverse systems. We prove that their construction and algebraic properties are independent of the chosen apolarity action. We propose a determinantal method for computing minimal local GADs by minimizing the rank of a symbolic inverse system. When the locus of minimal supports is finite, this provides a practical procedure to determine all minimal local decompositions without tensor extensions. We prove that this finiteness is guaranteed whenever the local GAD-rank of the form does not exceed its degree. We analyze both generic and special cases, provide computational evidence assessing the impact of different choices for minors, and compare our approach with existing algorithms for local apolar schemes. 2020 Mathematics Subject Classification Primary: 13H10, 14N07, 15A69 1. Introduction The study of additive decompositions of homogeneous polynomials is a central topic in commutative algebra and applied algebraic geometry, with far-reaching applications to symmetric tensor decomposition. The classical Waring decomposition problem , namely expressing a form as a minimal sum of powers of linear forms, has been extensively investigated in connection with secant varieties, catalecticant matrices, and apolarity theory [ 1 , 11 , 12 , 22 , 24 ] . From a scheme-theoretic perspective, the solution to the Waring problem for a given form F is equivalent to detecting minimal reduced zero-dimensional schemes that are apolar to F . Despite its theoretical interest and multitude of applications, this remains a highly challenging problem: computing such minimal configurations is, in general, difficult without additional structural assumptions [ 8 , 10 ] , even when such minima are unique ( identifiable ) [ 14 , 15 ] . Over the past decades, several generalizations of this problem have been considered, capturing more sophisticated intrinsic properties of a given form. The measures of such decompositions, known as their ranks [ 3 ] , reflect different aspects of the complexity of the considered form and are deeply intertwined with the structure of the associated apolar algebras. In particular, the Castelnuovo-Mumford regularity of such algebras has proven to be a key measure for the computational cost of effectively producing such minimal decompositions [ 2 , 7 , 18 , 25 ] . In this work, we tackle the problem of computing local generalized additive decompositions [ 22 ] , shortened as local GADs , of a given degree- d form F , namely decompositions of the form F=\omega\ell^{d-k}, where \omega and \ell are forms of degree k and 1 , respectively. Our goal is to effectively determine such decompositions with minimal algebraic complexity, measured by the length of the naturally associated apolar point scheme. Geometrically, this amounts to finding minimal representations of a projective point on the osculating varieties of the Veronese [ 2 , 4 ] . While the local Waring decomposition problem is trivial, its GAD counterpart is significantly richer. In fact, the linear span of non-reduced points displays substantially more varied geometric behavior, which is governed by the structure of (local) Artinian Gorenstein algebras and their inverse systems . These local objects are essential for understanding the cactus rank [ 11 ] , also known as minimal scheme length [ 22 ] , of forms. Indeed, these quantities can always be computed from GADs of certain primitives of F , and in the local case, these notions coincide with the size of a minimal local GAD when the socle degree of a minimal apolar scheme does not exceed the degree of F [ 7 ] . We currently have systematic methods to test and certify global minimality [ 2 , 8 ] , which have recently also been adapted to address the local case considered in this paper [ 5 ] . However, they rely on large and highly structured parameter spaces, making explicit computations rapidly infeasible even for moderately sized examples. Novel contributions We connect different notions of apolarity (differentiation, contraction, and \star -action) and their corresponding constructions of point schemes evinced by GADs, as they appear in recent literature [ 2 , 6 , 7 ] . In particular, we establish an explicit correspondence between the inverse systems defined by the respective dual generators ( Section 2.2 ), showing that it is compatible both with the derivation action (in the polynomial-exponential setting) and with the contraction action (in the divided power framework). We then adopt the latter viewpoint, and propose a new exact method to compute minimal local GADs of a given form in \Bbbk[x_{0},\dots,x_{n}] , which is based on minimizing the dimension of the corresponding symbolic inverse system ( Algorithm 1 ). This procedure is particularly effective when the set of minimal supports is finite, as we can collect determinantal relations from the inverse system matrix ( Section 3.1 ) of a form F at a symbolic linear form \ell\in\Bbbk[\gamma_{1},\dots,\gamma_{n}][x_{0},\dots,x_{n}] , until they define a zero-dimensional ideal in the parameter space \Bbbk[\gamma] . A key aspect in this process is the selection of suitable minors. When the size of a minimal local GAD F does not exceed its degree, we prove that a specific choice of minors implies the existence of a finite number of minimal supports ( section 3.1 ), which applies similarly for local cactus ( Section 3.1 ). For the general case, we propose and test different practical strategies, and we observe that the best performance is often achieved by following the contraction chains ( Section 3.4 ). Finally, we analyze the positive-dimensional loci of minimal supports for generic choices of the form F ( Section 3.3 ). We provide several examples where these results, jointly with the proposed algorithm, can be applied to understand the structure of every local GAD for certain given forms ( Sections 3.3 , 3.3 and 3.3 ). This analysis is much finer than what was achievable with previous methods, highlighting the effectiveness of the proposed approach for studying local GADs. Moreover, it does not involve any tensor extensions, and its complexity appears to rely only on the local GAD-rank and on the finiteness of the minimal loci, rather than the CM-regularity of the associated schemes. Paper organization The paper is organized as follows. In Section 2 , we review the necessary background on GADs, apolarity, and inverse systems. In Section 3 , we introduce the symbolic inverse system and apply it for understanding and computing minimal local GADs. In Section 4 , we compare our method with existing algorithms for local apolar schemes. Finally, we propose directions for future research in Section 5 . 2. Preliminaries 2.1. Notation In this paper, n\in\mathbb{N} will be a fixed positive integer, and \Bbbk will be a field. We assume that the characteristic of \Bbbk is zero or strictly larger than the degree of any form considered, and we will further assume that all the considered schemes \mathfrak{Z}\subset{\mathbb{P}}^{n}(\Bbbk) have \Bbbk -rational support. We will also employ the symbolic ring \mathbb{K}=\Bbbk[\gamma] , where \gamma=(\gamma_{1},\dots,\gamma_{n}) are parameters to be specialized in \Bbbk . We consider the standard graded multivariate polynomial \Bbbk -algebra \mathcal{S}=\oplus_{d\in\mathbb{N}}\mathcal{S}_{d}=\Bbbk[x_{0},\dots,x_{n}] and its differential ring \mathcal{R}=\oplus_{d\in\mathbb{N}}\mathcal{R}_{d}=\Bbbk[y_{0},\dots,y_{n}] , acting on \mathcal{S} by differentiation: y^{\beta}\circ x^{\alpha}=\begin{cases}\frac{\alpha!}{(\alpha-\beta)!}x^{\alpha-\beta}&\textnormal{if }\alpha\geq\beta,\\ 0&\textnormal{otherwise}.\end{cases}

Read the complete text