Loop-free Markov chains as determinantal point processes (Q731448)

A random subset \(X\) of a discrete space \(Y\) is considered. A distribution \(P\) of this subset satisfies the following condition: there exists a positive definite function \(K(x,y)\) \((x,y\in Y)\) such that for any \(n\geq2\), \(y_i\in Y\) it is fair \(P((y_1,\dots.y_n)\subset X)=\text{det}(K(y_i,y_j))_{n\times n}\). This random subset is said to be a determinantal point process with the correlation kernel \(K\). The author shows an example of such a point process. This is a homogeneous Markov chain with a transition matrix \(P_{xy}\) such that for any \(k\geq1\) and \(x\in Y\) \,\(P_{xx}^k=0\) (so called loop-free Markov chain). The family of realizations of this chain determines a random subset \(X\) of \(Y\), which is, by the author's assertion, a determinantal point process. The correlation kernel of this process is derived in terms of the original Markov chain. The author shows that renewal processes and semi-Markov processes on a discrete space can be considered as loop-free Markov chains. Some transformations of such a random subset are shown to preserve determinantal property of this subset. A variant of the central limit theorem for determinantal point processes is proved.