The threshold EM algorithm for parameter learning in bayesian network with incomplete data
This work addresses parameter learning issues in Bayesian networks for applications like medical diagnosis, but it appears incremental as it combines existing methods without a major breakthrough.
The paper tackles the problem of parameter learning in Bayesian networks with incomplete data by proposing a fusion of EM and RBE algorithms to limit search space and avoid local maxima. The method was applied to brain tumor diagnosis, showing advantages and disadvantages compared to the standard EM algorithm, but no concrete numerical results are provided.
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.