Well if my transfer matrix is unfeasibly large then I can’t construct the transfer matrix and so can not construct DiscreteMarkovChain, unless I misunderstood your suggestion. Imagine for instance a case where state space is 10 dimensional positive integer finite lattice up to 10000. State space has 10000*10 elements transfer matrix is to the power 2 of that…
The transfer matrix is used to construct the probability density of states at each step but since I can not do that, the only tool I have is to have a routine that generates individual trajectories from the markov chain and use these trajectories somehow…