A1198
Title: Bayesian nonparametric mixtures of Markov processes
Authors: Ramses Mena - Universidad Nacional Autonoma De Mexico (Mexico) [presenting]
Abstract: The purpose is to introduce a framework for constructing nonparametric Markov processes by starting from a base process with well-defined transition probabilities and invariant distributions. A random mixture mechanism is then used to generate new processes, where the dynamics are governed by random coefficients derived from the mixture structure. This approach yields a flexible class of Markov models that preserve key invariance properties while allowing for greater adaptability in modeling complex systems. To fully develop the framework, it is necessary to study fundamental theoretical aspects such as stability and symmetry. In addition, because the transition mechanism might involve infinitely many components, new strategies for estimation and prediction are required.
