Scalable Bayesian Learning of Recurrent Neural Networks for Language Modeling

ACL 2017

Abstract: Recurrent neural networks (RNNs) have shown promising performance for language modeling. However, traditional training of RNNs using back-propagation through time often suffers from overfitting. One reason for this is that stochastic optimization (used for large training sets) does not provide good estimates of model uncertainty. This paper leverages recent advances in stochastic gradient Markov Chain Monte Carlo (also appropriate for large training sets) to learn weight uncertainty in RNNs. It yields a principled Bayesian learning algorithm, adding gradient noise during training (enhancing exploration of the model-parameter space) and model averaging when testing. Extensive experiments on various RNN models and across a broad range of applications demonstrate the superiority of the proposed approach over stochastic optimization. Authors: Zhe Gan, Chunyuan Li, Changyou Chen, Yunchen Pu, Qinliang Su, Lawrence Carin (Duke University)