COBRA: Nonlinear Aggregation of Predictors
This package performs prediction for regression-oriented problems, aggregating in a nonlinear scheme any basic regression machines suggested by the context and provided by the user. If the user has no valuable knowledge on the data, four defaults machines wrappers are implemented so as to cover a minimal spectrum of prediction methods. If necessary, the computations may be parallelized. The method is described in Biau, Fischer, Guedj and Malley (2013), "COBRA: A Nonlinear Aggregation Strategy".