Learning Conditional Distributions Using Mixtures of Truncated Basis Functions

Abstract

Abstract. Mixtures of Truncated Basis Functions (MoTBFs) have recently been proposed for modelling univariate and joint distributions in hybrid Bayesian networks. In this paper we analyse the problem of learning conditional MoTBF distributions from data. Our approach utilizes a new technique for learning joint MoTBF densities, then propose a method for using these to generate the conditional distributions. The main contribution of this work is conveyed through an empirical investigation into the properties of the new learning procedure, where we also compare the merits of our approach to those obtained by other proposals.

Multiscalar genetic pathway modeling with hybrid Bayesian networks

“…Hybrid Bayesian networks are also an active area of research and several new hybrid Bayesian network methods have been recently proposed that allow for any relationships between discrete and continuous variables. These have included using mixtures of polynomials (Shenoy, ; Shenoy & West, ) and mixtures of truncated basis functions (Langseth, Nielsen, Perez‐Bernabe, & Salmeron, ; Langseth, Nielsen, Rumi, & Salmeron, ; Perez‐Bernabe, Salmeron, & Langseth, ) to approximate the distributions of the data. However, to a large extent, many of these methods have focused on parameter learning of the network (Salmeron, Rumi, Langseth, Nielsen, & Madsen, ).…”