Total: 1
We present an optimization-based formulation of the Red Light Green Light (RLGL) algorithm for computing stationary distributions of large Markov chains. This perspective clarifies the algorithm's behavior, establishes exponential convergence for a class of chains, and suggests practical scheduling strategies to accelerate convergence.