Just wanted to flag Lakshminarayanan et al. as a standard example of the “train ensemble with different initializations” approach
Just wanted to flag Lakshminarayanan et al. as a standard example of the “train ensemble with different initializations” approach