Tag
This paper establishes that one intervention per strongly connected component suffices to recover parameters of linear stochastic dynamics from steady-state data, and proposes a recursive learning algorithm.