integrativeME: integrative mixture of experts

Mixture of experts models (Jacobs et al., 1991) were introduced to account for nonlinearities and other complexities in the data. It is based on a divide-and-conquer strategy. Mixture of experts are of interest due to their wide applicability and the advantages of fast learning via the expectation-maximization (EM) algorithm. We have extended and implemented mixture of experts to combine categorical clinical factors and continuous microarray data in a binary classification framework to analyze cancer studies. To provide a hybrid signature of clinical factors and gene markers, we propose to apply different gene selection procedures as a first step.

Downloads: