Differentiable Scheduled Sampling for Credit Assignment

AbstractWe demonstrate that a continuous relaxation of the argmax operation can be used to create a differentiable approximation to greedy decoding in sequence-to-sequence (seq2seq) models. By incorporating this approximation into the scheduled sampling training procedure–a well-known technique for correcting exposure bias–we introduce a new training objective that is continuous and differentiable everywhere and can provide informative gradients near points where previous decoding decisions change their value. By using a related approximation, we also demonstrate a similar approach to sampled-based training. We show that our approach outperforms both standard cross-entropy training and scheduled sampling procedures in two sequence prediction tasks: named entity recognition and machine translation.