Bayesian Attention Belief Networks

              Shujian Zhang * 1 Xinjie Fan * 1 Bo Chen 2 Mingyuan Zhou 1

            Abstract                of the Transformer structure, it becomes possible to train
                                unprecedented large models on big datasets (Devlin et al.,
   Attention-based neural networks have achieved       2018), which stimulates a great amount of research to pre-
   state-of-the-art results on a wide range of tasks.    train models on large unlabeled d ...