Convergence of Markovian Stochastic Approximation with discontinuous dynamics

Stochastic approximation was introduced to find the roots of a deterministic function, often called the mean field, when only noisy measurements of it are available. In this article, we are interested in the convergence of such a method when the noise in the mean field measurement is Markov controlled and the dynamic (the function used to update the current parameter) is discontinuous. Two examples with discontinuous dynamics are given to illustrate our results, the first one being quantile estimation and the second one being vector quantization.