deep ensembles have recently gained popularity in the deep learning community
for their conceptual simplicity and efficiency. However, maintaining functional
diversity between ensemble members that are independently trained with gradient
descent is challenging. This can lead to patholo