PosteriorRegularization

original paper: http://www.jmlr.org/papers/volume11/ganchev10a/ganchev10a.pdf

• Prior Knowledge Integration for Neural Machine Translation using Posterior Regularization
• Harnessing Deep Neural Networks with Logic Rules
• Deep Neural Networks with Massive Learned Knowledge

Could be used for various reasons: to incorporate hand-curated features with end-to-end discriminative or semi-supervised training objectives; to include language model as a regularizer, etc.