Towards Robust Deep Learning using Entropic Losses

David Macêdo

Current deep learning solutions are well known for not informing whether they can reliably classify an example during inference. One of the most effective ways to build more reliab…