Embedding of Markov matrices for \(d \leqslant 4\) (Q6584149)

The classical embedding problem consists of finding a generator matrix \(Q\) (the infinitesimal operator of a continuous-time Markov chain) such that \(e^Q=M\), where \(M\) is a given Markov or stochastic matrix, representing the transition probabilities of a discrete-time Markov chain. In this paper, the authors address the case where the matrices are \(4\times4, (d=4)\) and provide a complete and simplified analysis of the case \(d = 3\). Additionally, they revisit the formulation of the embedding problem for time-inhomogeneous Markov chains.