Parameter Estimation in Mixtures of Truncated Exponentials

Bayesian networks with mixtures of truncated exponentials (MTEs) support efficient inference algorithms and provide a flexible way of modeling hybrid domains. On the other hand, estimating an MTE from data has turned out to be a difficult task, and most prevalent learning methods treat parameter estimation as a regression problem. The drawback of this approach is that by not directly attempting to find the parameters that maximize the likelihood, there is no principled way of e.g. performing subsequent model selection using those parameters. In this paper we describe an estimation method that directly aims at learning the maximum likelihood parameters of an MTE potential. Empirical results demonstrate that the proposed method yields significantly better likelihood results than regression-based methods.