Why is the use of deep ensembles for predictive uncertainty estimation considered both simple and scalable?
A. Deep ensembles are easy to implement and scale because they require only a single model to operate effectively.
B. Deep ensembles involve training multiple models and averaging their predictions, which provides a straightforward way to estimate uncertainty and can be scaled by increasing the number of models.
C. Deep ensembles are scalable but require complex algorithms and high computational power that makes them less accessible.
D. Deep ensembles are simple and scalable due to their reliance on traditional statistical methods rather than advanced machine learning techniques.