Efficient Discriminative Training Method for Structured Predictions

We propose an efficient discriminative training method for generative models under supervised learning. In our setting, fully observed instances are given as training examples, together with a specifi