The threshold EM algorithm for parameter learning in bayesian network with incomplete data

TL;DR

This paper introduces a threshold EM algorithm that combines EM and RBE methods for improved parameter learning in Bayesian networks with incomplete data, demonstrated through brain tumor diagnosis.

Contribution

It proposes a novel fusion of EM and RBE algorithms that incorporates parameter ranges to enhance learning in Bayesian networks with missing data.

Findings

Improved parameter estimation in Bayesian networks.

Application to brain tumor diagnosis shows advantages over standard EM.

Highlights limitations of the proposed method.

Abstract

Bayesian networks (BN) are used in a big range of applications but they have one issue concerning parameter learning. In real application, training data are always incomplete or some nodes are hidden. To deal with this problem many learning parameter algorithms are suggested foreground EM, Gibbs sampling and RBE algorithms. In order to limit the search space and escape from local maxima produced by executing EM algorithm, this paper presents a learning parameter algorithm that is a fusion of EM and RBE algorithms. This algorithm incorporates the range of a parameter into the EM algorithm. This range is calculated by the first step of RBE algorithm allowing a regularization of each parameter in bayesian network after the maximization step of the EM algorithm. The threshold EM algorithm is applied in brain tumor diagnosis and show some advantages and disadvantages over the EM algorithm.