Random monomial ideals (Q1628515): Difference between revisions

The authors define and study a random model on the set of monomial ideals of \(K[x_1,\dots,x_n]\). They call it the Erdős-Rényi-type model for random monomial ideals. Their definition follows closely the definition of \(\mathscr{G}(n,p)\), one of the two basic probability models of random graphs. See [\textit{B. Bollobás}, Random graphs. 2nd ed. Cambridge: Cambridge University Press (2001; Zbl 0979.05003)]. To be precise, the Erdős-Rényi-type model for random monomial ideals is a family with two parameters, \(0\leq p\leq 1\) and \(D\geq 0\), defined first by setting up a probability distribution on the sets of non-constant monomials of degree up to \(D\) and then using this distribution to induce a random model of monomial ideals. The probability distribution on the sets of monomial ideals is defined by stating that a monomial occurs in a random set -- denoted by \(\mathfrak{B}\) -- independently and with probability \(p\). Equivalently, given a set of non-constant monomials \(B\), \[ P(\mathfrak{B} = B) = p^{|B|}(1-p)^{\binom{D+n}{D}-|B|-1}. \] This distribution induces a random model of monomial ideals, denoted by \(\mathcal{I}(n,D,p)\), by positing \(\mathfrak{I} \sim \mathcal{I}(n,D,p)\) if \(\mathfrak{I} = (\mathfrak{B})\) where \(\mathfrak{B}\) is a random set of non-constant monomials of degree \(\leq D\), as constructed above. In Theorem 1.1, the authors compute the probability of a fixed monomial ideal \(I\) generated in degree \(\leq D\). The proof is as follows. The probability is equal to the probability of \(\mathfrak{B}\) (a random set of generators) to contain each element of the unique minimal set of monomial generators of \(I\) and to not contain any non-constant monomial of degree \(\leq D\) not in \(I\). By the definitions, these events are independent and can be decomposed into independent events. The probability that \(\mathfrak{B}\) contains each of the \(\beta_1=\beta_1(K[x_1,\dots,x_n]/I)\) elements of the minimal set of monomial generators is \(p^{\beta_1}\). The probability of it not containing any of the non-constant monomials of degree \(\leq D\) not in \(I\) is equal to \((1-p)^{\sum_{d=1}^D h(k)}\), where \(h\) is the Hilbert function of \(K[x_1,\dots,x_n]/I\). Thus: \[ P(\mathfrak{I} = I) = p^{\beta_1}(1-p)^{\sum_{d=1}^D h(k)}. \] In a similar vein, the authors compute the probability of and ideal \(I\) with a given Hilbert function (Theorem 2.4) and the probability of \(I\) yielding \(K[x_1,\dots,x_n]/I\) with a given Krull dimension (Theorems 3.1 and 3.2). Next, the asymptotic behavior of \(\mathfrak{I}\) is studied. This includes threshold functions for the Krull dimension of \(K[x_1,\dots,x_n]/\mathfrak{I}\) and the initial degree of \(\mathfrak{I}\). In the second to last section, a discussion of other possible probability models is carried out. Essentially, these are obtained by varying the probability of monomials in \(K[x_1,\dots,x_n]\). Using one such model, the authors show that random monomial ideals provide a generalization of the model of random simplicial complexes introduced in [\textit{A. Costa} and \textit{M. Farber}, J. Topol. Anal. 8, No. 3, 399--429 (2016; Zbl 1339.05440)]. The article finishes with a section of experimental data and conjectures. They concentrate on Cohen-Macaulayness, projective dimension, strong genericity, Castelnuovo-Mumford regularity and simplicial homology.