Tensor approach to mixed high-order moments of absorbing Markov chains (Q1940338)

Let \(N_j\) be a function giving the total number of times an absorbing Markov chain visits a transient state before absorption. The values of \(N_j\) depend on the state in which the Markov chain starts. Let \(E_i[N_j]\) be the first moment of \(N_j\) assuming the Markov chain starts at the state \(i\). The first moments and the non-mixed second moments can be easily expressed in matrix form using the fundamental matrix of the absorbing chain. However, the mixed second moments \(E_i[N_jN_k]\) and the mixed higher-order moments \(E_i[\prod_{j=0}^{m-1}N_{k_j}]\) have not been computed for an absorbing Markov chain in the general case. The paper solves this problem by using the tensor approach. It gives a brief introduction to the basic facts from tensor theory for this purpose, then determines the mixed second- and higher-order moments in tensor form, obtaining compact closed-form expressions.