Dynamic Sampling Procedure for Decomposable Random Networks

This research studies the problem of node ranking in a random network. Specifically, we consider a Markov chain with several ergodic classes and unknown transition probabilities which can be estimated by sampling. The objective is to select all of the best nodes in each ergodic class. A sampling procedure is proposed to decompose the Markov chain and maximize a weighted probability of correct selection of the best nodes in each ergodic class. Numerical results demonstrate the efficiency of the proposed sampling procedure.