Hybrid consensus theoretic classification with pruning and regularization

Conventional statistical pattern recognition methods are not appropriate in classification of multisource data since such data cannot, in most cases, be modeled by a common convenient multivariate statistical model. However, methods based on consensus theory have shown potential in classification of multisource data. Here, optimized combination, regularization, and pruning is proposed for consensus theoretic classification. The regularization scheme iteratively adapts regularization parameters by minimizing the validation error.