Finiteness for the \(k\)-factor model and chirality varieties (Q1043495)

This paper deals with two families of algebraic varieties arising from applications. First the \textit{\(k\)-factor model} in statistics, consisting of \(n\times n\) covariance matrices of \(n\) observed Gaussian random variables that are pairwise independent given \(k\) hidden Gaussian variables. Second, \textit{chirality varieties} inspired by applications in chemistry. A point in such chirality variety records chirality measurements of all \(k\)-subsets among an \(n\)-set of ligands. Both classes of varieties are given by a parameterisation, while for applications having polynomial equations is often desirable. The authors prove that in a precise sense, which is different for the two classes of varieties, these equations are finitely characterisable when \(k\) is fixed and \(n\) grows. Factor analysis addresses the problem of testing whether \(n\) observed random variables are conditionally independent given \(k\) hidden variables, called the factors. In the case where the joint distribution of all \(n+k\) variables is multivariate Gaussian, the parameter space \(F_{n,k}\) for the \(k\)-factor model is the set of \(n\times n\)-covariance matrices of the form \(\Sigma + \Gamma\) where \(\Sigma\) is diagonal positive definite and \(\Gamma\) is positive definite of rank at most \(k\). An algebraic approach to factor analysis, introduced by \textit{M. Drton}, \textit{B. Sturmfels}, and \textit{S. Sullivant} [Probab. Theory Related Fields 138, No. 3--4, 463--493 (2007; Zbl 1111.13020)] , seeks to determine all polynomial relations among the matrix entries in \(F_{n,k}\). Any principal \(m\times m\)-submatrix of a matrix in \(F_{n,k}\) lies in \(F_{m,k}\). An important question of theoretical interest is whether, for fixed \(k\), there exists an \(m\) such that for \(n \geq m\) the model \(F_{n,k}\) is completely characterised by the fact that each principal \(m\times m\)-submatrix lies in \(F_{k,m}\). For \(k=2\) this question was settled in the affirmative (\(m = 6\) suffices) by \textit{M. Drton} and \textit{H. Xiao} [Ann. Inst. Stat. Math. 62, No. 4, 775--783 (2010)]. In this paper, Draisma proves the corresponding statement for the Zariski closure of the model, i.e., for the set of all real \(n\times n\)-matrices satisfying all model invariants. This result settles the ``radical'' part of Question 29 in the paper by Drton, Sturmfels, and Sullivant cited above. The second finiteness result concerns another family of algebraic varieties, motivated by applications in chemistry. Fix a natural number \(k\) and for all \(n \geq k\) consider the affine space \(S_{n}^{(k)}\) whose coordinates \(y_{J}\) are indexed by all \(k\)-subsets of \([n]\). Consider the morphism \(\mathbb{A}^{n} \to S_{n}^{(k)}\) that sends a point \(x\) to the point \(y\) in \(S_{n}^{(k)}\) whose coordinate \(y_{J}\) equals the Vandermonde determinant \(\prod_{i,j \in J: i<j} (x_{i} - x_{j})\). The Zariski closure of the image of this map is called the \textit{chirality variety} \(V_{n}^{(k)}\). In this paper, Draisma proves that there is a natural number \(N\), depending only on \(k\), such that for all \(n \geq N\), the ideal of \(V_{n}^{(k)}\) is generated by the pull-backs of \(V_{N}^{(k)}\) under all projections \(\pi_{I}: S_{n}^{(k)} \to S_{N}^{(k)}\), where \(I\) ranges over all \(N\)-subsets of \([n]\). To prove these finiteness results, Draisma first establishes a common framework in which a sequence of schemes with group actions is replaced by its limit, which is a scheme of infinite type. Then he introduces and develops the new notion of \(G\)-Noetherianity for topological spaces and also for rings. This notion is based on the seminal works on finite generation of ideals in infinite dimensional polynomial rings by \textit{M. Aschenbrenner} and \textit{C. Hillar} [Trans. Am. Math. Soc. 359, No. 11, 5171--5192 (2007); erratum ibid. 361, No. 10, 5627--5627 (2009; Zbl 1129.13008)] and \textit{C. J. Hillar} and \textit{S. Sullivant} [Adv. Math. 229, No. 1, 1--25 (2012; Zbl 1233.13012)].