Equivariant Hilbert series for hierarchical models (Q2032062)

The paper under review studies the construction of equivariant Hilbert series for hierarchical models. These models are used in algebraic statistics to determine dependencies among random variables. A hierarchical model \({\mathcal M}={\mathcal M}(\Delta,\mathbf{r})\) with \(m\) parameters is given by a collection \(\Delta=\{F_1,\ldots,F_q\}\) of nonempty subsets \(F_j\subset \{1,\ldots ,m\}\) with \(\bigcup_{j=1}^{q}F_j=\{1,\ldots ,m\}\) and a sequence \(\mathbf{r}=(r_1,\ldots ,r_m)\) of \(m\) positive integers. Each parameter corresponds to a random variable, and \(r_i\) denotes the number of values parameter \(i\) can take. Furthermore, the toric ideal associated to this hierarchical model is denoted by \(I_{\mathcal{M}}\). The authors, study the symmetry of \(I_{\mathcal{M}}\), its invariant filtration as well as it equivariant Hilbert series. They give a condition that guarantees that the equivariant Hilbert series is a rational function in \(m+1\) variables with rational coefficients. To determine the coefficients of this function, they benefit from the construction of finite automata that recognize languages corresponding to invariant filtrations.