Combining Protein Secondary Structure Prediction Methods with a new Multi-Category SVM
Résumé
Building upon a new uniform convergence result derived for real-valued multi-category discriminant models, we propose a theoretical framework for multi-category SVMs design. One of the new SVMs specified in that way is assessed on an open problem in predictive structural biology: protein secondary structure prediction. It is used as an ensemble method to combine the outputs of three of the current best prediction methods.