Two-level Mixtures of Markov Trees

Bayesian Networks efficiently encode a probability distribution on a large set of variables but their poor scaling in terms of the number of variables may make them unfit to tackle learning and inference problems of increasing size. Mixtures of Markov trees however scale well by design. We investigate whether the two approaches to learn such mixtures from data (maximum likelihood and variance reduction) can be combined together by building a two level Mixture of Markov Trees. Our experiments on synthetic data show the interest of this model.