Weighted committee-based structure learning for microarray data

Bayesian networks (BN) are considered to be one of the strongest modeling techniques of gene regulatory networks (GRN) thanks to their ability to present features and relations between them in a causal and probabilistic way. Learning the structure of those models needs a large training dataset in order to avoid over-fitting. However, biological data, especially microarray data, suffer from the presence of only few instances. Some recent approaches tried to face this challenge by applying committee based methods. We use this principle in order to suggest a new method supported by a double-weight-assignment technique. We show that our approach has succeeded to learn benchmark structures.